| Message ID | 1648103831-12347-4-git-send-email-quic_c_sanm@quicinc.com (mailing list archive) |
|---|---|
| State | Not Applicable |
| Headers | show |
| Series | Refactor xhci quirks and plat private data | expand |
On Thu, Mar 24, 2022 at 12:07:11PM +0530, Sandeep Maheswaram wrote: > Currently the phy init is done from dwc3 and also xhci which makes the > runtime_usage value 2 for the phy which causes issue during runtime > suspend. When we run the below command the runtime_status still shows > active. > echo auto > /sys/bus/platform/devices/88e3000.phy/power/control > > dwc3 manages PHY by own DRD driver, so skip the management by > HCD core by setting this quirk. > > Signed-off-by: Sandeep Maheswaram <quic_c_sanm@quicinc.com> > --- > drivers/usb/dwc3/host.c | 13 +++++++++++++ > 1 file changed, 13 insertions(+) > > diff --git a/drivers/usb/dwc3/host.c b/drivers/usb/dwc3/host.c > index eda8719..d4fcf06 100644 > --- a/drivers/usb/dwc3/host.c > +++ b/drivers/usb/dwc3/host.c > @@ -13,6 +13,12 @@ > #include <linux/platform_device.h> > > #include "core.h" > +#include <linux/usb/xhci-plat.h> > +#include <linux/usb/xhci-quirks.h> > + > +static const struct xhci_plat_priv xhci_plat_dwc3_xhci = { > + .quirks = XHCI_SKIP_PHY_INIT, > +}; > > static void dwc3_host_fill_xhci_irq_res(struct dwc3 *dwc, > int irq, char *name) > @@ -122,6 +128,13 @@ int dwc3_host_init(struct dwc3 *dwc) > } > } > > + ret = platform_device_add_data(xhci, &xhci_plat_dwc3_xhci, > + sizeof(xhci_plat_dwc3_xhci)); > + 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"); I think you should just use device property: diff --git a/drivers/usb/dwc3/host.c b/drivers/usb/dwc3/host.c index eda871973d6cc..dbff7b8ed6d5e 100644 --- a/drivers/usb/dwc3/host.c +++ b/drivers/usb/dwc3/host.c @@ -67,7 +67,7 @@ static int dwc3_host_get_irq(struct dwc3 *dwc) int dwc3_host_init(struct dwc3 *dwc) { - struct property_entry props[4]; + struct property_entry props[5]; struct platform_device *xhci; int ret, irq; int prop_idx = 0; @@ -114,12 +114,12 @@ int dwc3_host_init(struct dwc3 *dwc) if (DWC3_VER_IS_WITHIN(DWC3, ANY, 300A)) props[prop_idx++] = PROPERTY_ENTRY_BOOL("quirk-broken-port-ped"); - if (prop_idx) { - ret = device_create_managed_software_node(&xhci->dev, props, NULL); - if (ret) { - dev_err(dwc->dev, "failed to add properties to xHCI\n"); - goto err; - } + props[prop_idx++] = PROPERTY_ENTRY_BOOL("skip-phy-init"); + + ret = device_create_managed_software_node(&xhci->dev, props, NULL); + if (ret) { + dev_err(dwc->dev, "failed to add properties to xHCI\n"); + goto err; } ret = platform_device_add(xhci); diff --git a/drivers/usb/host/xhci-plat.c b/drivers/usb/host/xhci-plat.c index 649ffd861b44e..31ed39d06e29b 100644 --- a/drivers/usb/host/xhci-plat.c +++ b/drivers/usb/host/xhci-plat.c @@ -307,6 +307,9 @@ static int xhci_plat_probe(struct platform_device *pdev) device_property_read_u32(tmpdev, "imod-interval-ns", &xhci->imod_interval); + + if (device_property_read_bool(tmpdev, "skip-phy-init")) + hcd->skip_phy_initialization = 1; } hcd->usb_phy = devm_usb_get_phy_by_phandle(sysdev, "usb-phy", 0);
On 24.3.2022 14.27, Heikki Krogerus wrote: > On Thu, Mar 24, 2022 at 12:07:11PM +0530, Sandeep Maheswaram wrote: >> Currently the phy init is done from dwc3 and also xhci which makes the >> runtime_usage value 2 for the phy which causes issue during runtime >> suspend. When we run the below command the runtime_status still shows >> active. >> echo auto > /sys/bus/platform/devices/88e3000.phy/power/control >> >> dwc3 manages PHY by own DRD driver, so skip the management by >> HCD core by setting this quirk. >> >> Signed-off-by: Sandeep Maheswaram <quic_c_sanm@quicinc.com> >> --- >> drivers/usb/dwc3/host.c | 13 +++++++++++++ >> 1 file changed, 13 insertions(+) >> >> diff --git a/drivers/usb/dwc3/host.c b/drivers/usb/dwc3/host.c >> index eda8719..d4fcf06 100644 >> --- a/drivers/usb/dwc3/host.c >> +++ b/drivers/usb/dwc3/host.c >> @@ -13,6 +13,12 @@ >> #include <linux/platform_device.h> >> >> #include "core.h" >> +#include <linux/usb/xhci-plat.h> >> +#include <linux/usb/xhci-quirks.h> >> + >> +static const struct xhci_plat_priv xhci_plat_dwc3_xhci = { >> + .quirks = XHCI_SKIP_PHY_INIT, >> +}; >> >> static void dwc3_host_fill_xhci_irq_res(struct dwc3 *dwc, >> int irq, char *name) >> @@ -122,6 +128,13 @@ int dwc3_host_init(struct dwc3 *dwc) >> } >> } >> >> + ret = platform_device_add_data(xhci, &xhci_plat_dwc3_xhci, >> + sizeof(xhci_plat_dwc3_xhci)); >> + 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"); > > I think you should just use device property: > This was suggested in an earlier series, but was rejected as it also added the property as a device tree parameter. I think adding more device properties can be messy in the long run, especially if we need to add them for many of the existing xhci quirks. We also end up with a mix where some device properties are listed as device tree parameters, and some not. Defining xhci quirks and platform data structure in headers shared with dwc3 and cdns3 allow those drivers to easily set any existing xhci quirk, or other possible optional callbacks. cdns3 driver is already doing this, but it includes the full xhci.h header. This series cleans up that a bit so cdns3 will only include xhci quirk bits and platform data structure. On the downside we add a couple xhci related header files to include/linux/usb/ Let me know if you see any other issues I missed with this approach. Thanks Mathias
On Fri, Mar 25, 2022 at 12:36:22AM +0200, Mathias Nyman wrote: > On 24.3.2022 14.27, Heikki Krogerus wrote: > > On Thu, Mar 24, 2022 at 12:07:11PM +0530, Sandeep Maheswaram wrote: > >> Currently the phy init is done from dwc3 and also xhci which makes the > >> runtime_usage value 2 for the phy which causes issue during runtime > >> suspend. When we run the below command the runtime_status still shows > >> active. > >> echo auto > /sys/bus/platform/devices/88e3000.phy/power/control > >> > >> dwc3 manages PHY by own DRD driver, so skip the management by > >> HCD core by setting this quirk. > >> > >> Signed-off-by: Sandeep Maheswaram <quic_c_sanm@quicinc.com> > >> --- > >> drivers/usb/dwc3/host.c | 13 +++++++++++++ > >> 1 file changed, 13 insertions(+) > >> > >> diff --git a/drivers/usb/dwc3/host.c b/drivers/usb/dwc3/host.c > >> index eda8719..d4fcf06 100644 > >> --- a/drivers/usb/dwc3/host.c > >> +++ b/drivers/usb/dwc3/host.c > >> @@ -13,6 +13,12 @@ > >> #include <linux/platform_device.h> > >> > >> #include "core.h" > >> +#include <linux/usb/xhci-plat.h> > >> +#include <linux/usb/xhci-quirks.h> > >> + > >> +static const struct xhci_plat_priv xhci_plat_dwc3_xhci = { > >> + .quirks = XHCI_SKIP_PHY_INIT, > >> +}; > >> > >> static void dwc3_host_fill_xhci_irq_res(struct dwc3 *dwc, > >> int irq, char *name) > >> @@ -122,6 +128,13 @@ int dwc3_host_init(struct dwc3 *dwc) > >> } > >> } > >> > >> + ret = platform_device_add_data(xhci, &xhci_plat_dwc3_xhci, > >> + sizeof(xhci_plat_dwc3_xhci)); > >> + 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"); > > > > I think you should just use device property: > > > > This was suggested in an earlier series, but was rejected as it also added > the property as a device tree parameter. > > I think adding more device properties can be messy in the long run, especially if we > need to add them for many of the existing xhci quirks. > We also end up with a mix where some device properties are listed as device tree > parameters, and some not. > > Defining xhci quirks and platform data structure in headers shared with dwc3 and cdns3 > allow those drivers to easily set any existing xhci quirk, or other possible optional > callbacks. > > cdns3 driver is already doing this, but it includes the full xhci.h header. > This series cleans up that a bit so cdns3 will only include xhci quirk bits and > platform data structure. > > On the downside we add a couple xhci related header files to include/linux/usb/ > Let me know if you see any other issues I missed with this approach. The problem here is that these drivers are now coupled together, and that should not be taken lightly. We have a dependency hell in our hands with a lot of drivers, and the culprit is always platform data. Build-in device properties may be messy, but I would still say they are less messy than those quirk flags - you got to admit, they are a mess. The benefit from build-in properties is in any case the fact that they remove the need to couple these drivers together. You can also use something like naming convention if you are worried about confusion between devicetree properties and build-in only properties ("build-in:skip-phy-init" or whatever), and of course require that each of the build-in only property is documented clearly in drivers/usb/host/xhci-plat.c. But this in any case really can not be justification for a platform data blob just so you can avoid using the properties - honestly, it really should to be the other way around. Platform data is in practice always problematic. On top of the driver coupling, it creates maintenance burden, code duplication, etc. Please don't just accept it lightly. I'm telling you, for hacks like this, the build-in device properties is a much much safer bet. thanks,
On 25.3.2022 13.27, Heikki Krogerus wrote: > On Fri, Mar 25, 2022 at 12:36:22AM +0200, Mathias Nyman wrote: >> On 24.3.2022 14.27, Heikki Krogerus wrote: >>> On Thu, Mar 24, 2022 at 12:07:11PM +0530, Sandeep Maheswaram wrote: >>>> Currently the phy init is done from dwc3 and also xhci which makes the >>>> runtime_usage value 2 for the phy which causes issue during runtime >>>> suspend. When we run the below command the runtime_status still shows >>>> active. >>>> echo auto > /sys/bus/platform/devices/88e3000.phy/power/control >>>> >>>> dwc3 manages PHY by own DRD driver, so skip the management by >>>> HCD core by setting this quirk. >>>> >>>> Signed-off-by: Sandeep Maheswaram <quic_c_sanm@quicinc.com> >>>> --- >>>> drivers/usb/dwc3/host.c | 13 +++++++++++++ >>>> 1 file changed, 13 insertions(+) >>>> >>>> diff --git a/drivers/usb/dwc3/host.c b/drivers/usb/dwc3/host.c >>>> index eda8719..d4fcf06 100644 >>>> --- a/drivers/usb/dwc3/host.c >>>> +++ b/drivers/usb/dwc3/host.c >>>> @@ -13,6 +13,12 @@ >>>> #include <linux/platform_device.h> >>>> >>>> #include "core.h" >>>> +#include <linux/usb/xhci-plat.h> >>>> +#include <linux/usb/xhci-quirks.h> >>>> + >>>> +static const struct xhci_plat_priv xhci_plat_dwc3_xhci = { >>>> + .quirks = XHCI_SKIP_PHY_INIT, >>>> +}; >>>> >>>> static void dwc3_host_fill_xhci_irq_res(struct dwc3 *dwc, >>>> int irq, char *name) >>>> @@ -122,6 +128,13 @@ int dwc3_host_init(struct dwc3 *dwc) >>>> } >>>> } >>>> >>>> + ret = platform_device_add_data(xhci, &xhci_plat_dwc3_xhci, >>>> + sizeof(xhci_plat_dwc3_xhci)); >>>> + 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"); >>> >>> I think you should just use device property: >>> >> >> This was suggested in an earlier series, but was rejected as it also added >> the property as a device tree parameter. >> >> I think adding more device properties can be messy in the long run, especially if we >> need to add them for many of the existing xhci quirks. >> We also end up with a mix where some device properties are listed as device tree >> parameters, and some not. >> >> Defining xhci quirks and platform data structure in headers shared with dwc3 and cdns3 >> allow those drivers to easily set any existing xhci quirk, or other possible optional >> callbacks. >> >> cdns3 driver is already doing this, but it includes the full xhci.h header. >> This series cleans up that a bit so cdns3 will only include xhci quirk bits and >> platform data structure. >> >> On the downside we add a couple xhci related header files to include/linux/usb/ >> Let me know if you see any other issues I missed with this approach. > > The problem here is that these drivers are now coupled together, and > that should not be taken lightly. We have a dependency hell in our > hands with a lot of drivers, and the culprit is always platform data. > > Build-in device properties may be messy, but I would still say they > are less messy than those quirk flags - you got to admit, they are a > mess. The benefit from build-in properties is in any case the fact > that they remove the need to couple these drivers together. Agree, quirk bits are messy. Any suggestion that would work with PCI xHCI devices, devicetree, and "pure" platform devices? > > You can also use something like naming convention if you are worried > about confusion between devicetree properties and build-in only > properties ("build-in:skip-phy-init" or whatever), and of course > require that each of the build-in only property is documented clearly > in drivers/usb/host/xhci-plat.c. But this in any case really can not > be justification for a platform data blob just so you can avoid using > the properties - honestly, it really should to be the other way > around. > > Platform data is in practice always problematic. On top of the driver > coupling, it creates maintenance burden, code duplication, etc. Please > don't just accept it lightly. I'm telling you, for hacks like this, the > build-in device properties is a much much safer bet. > Fair enough. Any idea if there's a long term solution for platform devices? Adding some type of "compatibility" id to platform devices in addition to name? Also just noticed cdns3 driver passes a function pointer via platform_data to xhci-plat. This should be cleared up as well. Thanks -Mathias
On Fri, Mar 25, 2022 at 04:33:27PM +0200, Mathias Nyman wrote: > On 25.3.2022 13.27, Heikki Krogerus wrote: > > On Fri, Mar 25, 2022 at 12:36:22AM +0200, Mathias Nyman wrote: > >> On 24.3.2022 14.27, Heikki Krogerus wrote: > >>> On Thu, Mar 24, 2022 at 12:07:11PM +0530, Sandeep Maheswaram wrote: > >>>> Currently the phy init is done from dwc3 and also xhci which makes the > >>>> runtime_usage value 2 for the phy which causes issue during runtime > >>>> suspend. When we run the below command the runtime_status still shows > >>>> active. > >>>> echo auto > /sys/bus/platform/devices/88e3000.phy/power/control > >>>> > >>>> dwc3 manages PHY by own DRD driver, so skip the management by > >>>> HCD core by setting this quirk. > >>>> > >>>> Signed-off-by: Sandeep Maheswaram <quic_c_sanm@quicinc.com> > >>>> --- > >>>> drivers/usb/dwc3/host.c | 13 +++++++++++++ > >>>> 1 file changed, 13 insertions(+) > >>>> > >>>> diff --git a/drivers/usb/dwc3/host.c b/drivers/usb/dwc3/host.c > >>>> index eda8719..d4fcf06 100644 > >>>> --- a/drivers/usb/dwc3/host.c > >>>> +++ b/drivers/usb/dwc3/host.c > >>>> @@ -13,6 +13,12 @@ > >>>> #include <linux/platform_device.h> > >>>> > >>>> #include "core.h" > >>>> +#include <linux/usb/xhci-plat.h> > >>>> +#include <linux/usb/xhci-quirks.h> > >>>> + > >>>> +static const struct xhci_plat_priv xhci_plat_dwc3_xhci = { > >>>> + .quirks = XHCI_SKIP_PHY_INIT, > >>>> +}; > >>>> > >>>> static void dwc3_host_fill_xhci_irq_res(struct dwc3 *dwc, > >>>> int irq, char *name) > >>>> @@ -122,6 +128,13 @@ int dwc3_host_init(struct dwc3 *dwc) > >>>> } > >>>> } > >>>> > >>>> + ret = platform_device_add_data(xhci, &xhci_plat_dwc3_xhci, > >>>> + sizeof(xhci_plat_dwc3_xhci)); > >>>> + 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"); > >>> > >>> I think you should just use device property: > >>> > >> > >> This was suggested in an earlier series, but was rejected as it also added > >> the property as a device tree parameter. > >> > >> I think adding more device properties can be messy in the long run, especially if we > >> need to add them for many of the existing xhci quirks. > >> We also end up with a mix where some device properties are listed as device tree > >> parameters, and some not. > >> > >> Defining xhci quirks and platform data structure in headers shared with dwc3 and cdns3 > >> allow those drivers to easily set any existing xhci quirk, or other possible optional > >> callbacks. > >> > >> cdns3 driver is already doing this, but it includes the full xhci.h header. > >> This series cleans up that a bit so cdns3 will only include xhci quirk bits and > >> platform data structure. > >> > >> On the downside we add a couple xhci related header files to include/linux/usb/ > >> Let me know if you see any other issues I missed with this approach. > > > > The problem here is that these drivers are now coupled together, and > > that should not be taken lightly. We have a dependency hell in our > > hands with a lot of drivers, and the culprit is always platform data. > > > > Build-in device properties may be messy, but I would still say they > > are less messy than those quirk flags - you got to admit, they are a > > mess. The benefit from build-in properties is in any case the fact > > that they remove the need to couple these drivers together. > > Agree, quirk bits are messy. Any suggestion that would work with > PCI xHCI devices, devicetree, and "pure" platform devices? I think xHCI driver should always be able to rely on being able to read this kind of information from the fwnode. If there is no actual firmware node (DT or ACPI), or if it's missing some information, the glue driver needs to populate software node for the xHCI. Right now I just want to avoid having to pass the quirks using platform data from drivers such as drivers/usb/cdns3/host.c and drivers/usb/dwc3/host.c to xHCI. One way we could do that is by defining compatibility ID for both of them that we provide using a single device property (like I guess DT does). Then based on that compatibility ID, xhci-plat.c can set the actual "static" quirk flags. That we could already do easily. How would that sound to you? > > You can also use something like naming convention if you are worried > > about confusion between devicetree properties and build-in only > > properties ("build-in:skip-phy-init" or whatever), and of course > > require that each of the build-in only property is documented clearly > > in drivers/usb/host/xhci-plat.c. But this in any case really can not > > be justification for a platform data blob just so you can avoid using > > the properties - honestly, it really should to be the other way > > around. > > > > Platform data is in practice always problematic. On top of the driver > > coupling, it creates maintenance burden, code duplication, etc. Please > > don't just accept it lightly. I'm telling you, for hacks like this, the > > build-in device properties is a much much safer bet. > > > > Fair enough. > > Any idea if there's a long term solution for platform devices? > Adding some type of "compatibility" id to platform devices in addition to name? We don't need to touch the platform devices if, and when, we have always the fwnode. The "compatibility" id should always be possible to pick from the fwnode. > Also just noticed cdns3 driver passes a function pointer via platform_data to > xhci-plat. This should be cleared up as well. Looks like a PM quirk. For that I would propose this: https://lore.kernel.org/linux-acpi/20201029105941.63410-2-heikki.krogerus@linux.intel.com/ So with that you can supply separate PM operations for your software fwnode. When ever the device suspends, the suspend quirk is also executed. xHCI core does not need to know about it. Yeah, I guess I need to finish that series... thanks,
Hi Mathias,Heikki On 3/25/2022 9:08 PM, Heikki Krogerus wrote: > On Fri, Mar 25, 2022 at 04:33:27PM +0200, Mathias Nyman wrote: >> On 25.3.2022 13.27, Heikki Krogerus wrote: >>> On Fri, Mar 25, 2022 at 12:36:22AM +0200, Mathias Nyman wrote: >>>> On 24.3.2022 14.27, Heikki Krogerus wrote: >>>>> On Thu, Mar 24, 2022 at 12:07:11PM +0530, Sandeep Maheswaram wrote: >>>>>> Currently the phy init is done from dwc3 and also xhci which makes the >>>>>> runtime_usage value 2 for the phy which causes issue during runtime >>>>>> suspend. When we run the below command the runtime_status still shows >>>>>> active. >>>>>> echo auto > /sys/bus/platform/devices/88e3000.phy/power/control >>>>>> >>>>>> dwc3 manages PHY by own DRD driver, so skip the management by >>>>>> HCD core by setting this quirk. >>>>>> >>>>>> Signed-off-by: Sandeep Maheswaram <quic_c_sanm@quicinc.com> >>>>>> --- >>>>>> drivers/usb/dwc3/host.c | 13 +++++++++++++ >>>>>> 1 file changed, 13 insertions(+) >>>>>> >>>>>> diff --git a/drivers/usb/dwc3/host.c b/drivers/usb/dwc3/host.c >>>>>> index eda8719..d4fcf06 100644 >>>>>> --- a/drivers/usb/dwc3/host.c >>>>>> +++ b/drivers/usb/dwc3/host.c >>>>>> @@ -13,6 +13,12 @@ >>>>>> #include <linux/platform_device.h> >>>>>> >>>>>> #include "core.h" >>>>>> +#include <linux/usb/xhci-plat.h> >>>>>> +#include <linux/usb/xhci-quirks.h> >>>>>> + >>>>>> +static const struct xhci_plat_priv xhci_plat_dwc3_xhci = { >>>>>> + .quirks = XHCI_SKIP_PHY_INIT, >>>>>> +}; >>>>>> >>>>>> static void dwc3_host_fill_xhci_irq_res(struct dwc3 *dwc, >>>>>> int irq, char *name) >>>>>> @@ -122,6 +128,13 @@ int dwc3_host_init(struct dwc3 *dwc) >>>>>> } >>>>>> } >>>>>> >>>>>> + ret = platform_device_add_data(xhci, &xhci_plat_dwc3_xhci, >>>>>> + sizeof(xhci_plat_dwc3_xhci)); >>>>>> + 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"); >>>>> I think you should just use device property: >>>>> >>>> This was suggested in an earlier series, but was rejected as it also added >>>> the property as a device tree parameter. >>>> >>>> I think adding more device properties can be messy in the long run, especially if we >>>> need to add them for many of the existing xhci quirks. >>>> We also end up with a mix where some device properties are listed as device tree >>>> parameters, and some not. >>>> >>>> Defining xhci quirks and platform data structure in headers shared with dwc3 and cdns3 >>>> allow those drivers to easily set any existing xhci quirk, or other possible optional >>>> callbacks. >>>> >>>> cdns3 driver is already doing this, but it includes the full xhci.h header. >>>> This series cleans up that a bit so cdns3 will only include xhci quirk bits and >>>> platform data structure. >>>> >>>> On the downside we add a couple xhci related header files to include/linux/usb/ >>>> Let me know if you see any other issues I missed with this approach. >>> The problem here is that these drivers are now coupled together, and >>> that should not be taken lightly. We have a dependency hell in our >>> hands with a lot of drivers, and the culprit is always platform data. >>> >>> Build-in device properties may be messy, but I would still say they >>> are less messy than those quirk flags - you got to admit, they are a >>> mess. The benefit from build-in properties is in any case the fact >>> that they remove the need to couple these drivers together. >> Agree, quirk bits are messy. Any suggestion that would work with >> PCI xHCI devices, devicetree, and "pure" platform devices? > I think xHCI driver should always be able to rely on being able to > read this kind of information from the fwnode. If there is no actual > firmware node (DT or ACPI), or if it's missing some information, the > glue driver needs to populate software node for the xHCI. > > Right now I just want to avoid having to pass the quirks using > platform data from drivers such as drivers/usb/cdns3/host.c and > drivers/usb/dwc3/host.c to xHCI. > > One way we could do that is by defining compatibility ID for both of > them that we provide using a single device property (like I guess DT > does). Then based on that compatibility ID, xhci-plat.c can set the > actual "static" quirk flags. That we could already do easily. How > would that sound to you? This was my previous patch where I was using device tree property. Should we go ahead with this approach? https://patchwork.kernel.org/project/linux-arm-msm/cover/1636353710-25582-1-git-send-email-quic_c_sanm@quicinc.com/ Any further changes to this ? Regards Sandeep > >>> You can also use something like naming convention if you are worried >>> about confusion between devicetree properties and build-in only >>> properties ("build-in:skip-phy-init" or whatever), and of course >>> require that each of the build-in only property is documented clearly >>> in drivers/usb/host/xhci-plat.c. But this in any case really can not >>> be justification for a platform data blob just so you can avoid using >>> the properties - honestly, it really should to be the other way >>> around. >>> >>> Platform data is in practice always problematic. On top of the driver >>> coupling, it creates maintenance burden, code duplication, etc. Please >>> don't just accept it lightly. I'm telling you, for hacks like this, the >>> build-in device properties is a much much safer bet. >>> >> Fair enough. >> >> Any idea if there's a long term solution for platform devices? >> Adding some type of "compatibility" id to platform devices in addition to name? > We don't need to touch the platform devices if, and when, we have > always the fwnode. The "compatibility" id should always be possible to > pick from the fwnode. > >> Also just noticed cdns3 driver passes a function pointer via platform_data to >> xhci-plat. This should be cleared up as well. > Looks like a PM quirk. For that I would propose this: > https://lore.kernel.org/linux-acpi/20201029105941.63410-2-heikki.krogerus@linux.intel.com/ > > So with that you can supply separate PM operations for your software > fwnode. When ever the device suspends, the suspend quirk is also > executed. xHCI core does not need to know about it. > > Yeah, I guess I need to finish that series... > > thanks, >
On Tue, Mar 29, 2022 at 02:48:05PM +0530, Sandeep Maheswaram (Temp) wrote: > Hi Mathias,Heikki > > On 3/25/2022 9:08 PM, Heikki Krogerus wrote: > >On Fri, Mar 25, 2022 at 04:33:27PM +0200, Mathias Nyman wrote: > >>On 25.3.2022 13.27, Heikki Krogerus wrote: > >>>On Fri, Mar 25, 2022 at 12:36:22AM +0200, Mathias Nyman wrote: > >>>>On 24.3.2022 14.27, Heikki Krogerus wrote: > >>>>>On Thu, Mar 24, 2022 at 12:07:11PM +0530, Sandeep Maheswaram wrote: > >>>>>>Currently the phy init is done from dwc3 and also xhci which makes the > >>>>>>runtime_usage value 2 for the phy which causes issue during runtime > >>>>>>suspend. When we run the below command the runtime_status still shows > >>>>>>active. > >>>>>>echo auto > /sys/bus/platform/devices/88e3000.phy/power/control > >>>>>> > >>>>>>dwc3 manages PHY by own DRD driver, so skip the management by > >>>>>>HCD core by setting this quirk. > >>>>>> > >>>>>>Signed-off-by: Sandeep Maheswaram <quic_c_sanm@quicinc.com> > >>>>>>--- > >>>>>> drivers/usb/dwc3/host.c | 13 +++++++++++++ > >>>>>> 1 file changed, 13 insertions(+) > >>>>>> > >>>>>>diff --git a/drivers/usb/dwc3/host.c b/drivers/usb/dwc3/host.c > >>>>>>index eda8719..d4fcf06 100644 > >>>>>>--- a/drivers/usb/dwc3/host.c > >>>>>>+++ b/drivers/usb/dwc3/host.c > >>>>>>@@ -13,6 +13,12 @@ > >>>>>> #include <linux/platform_device.h> > >>>>>> #include "core.h" > >>>>>>+#include <linux/usb/xhci-plat.h> > >>>>>>+#include <linux/usb/xhci-quirks.h> > >>>>>>+ > >>>>>>+static const struct xhci_plat_priv xhci_plat_dwc3_xhci = { > >>>>>>+ .quirks = XHCI_SKIP_PHY_INIT, > >>>>>>+}; > >>>>>> static void dwc3_host_fill_xhci_irq_res(struct dwc3 *dwc, > >>>>>> int irq, char *name) > >>>>>>@@ -122,6 +128,13 @@ int dwc3_host_init(struct dwc3 *dwc) > >>>>>> } > >>>>>> } > >>>>>>+ ret = platform_device_add_data(xhci, &xhci_plat_dwc3_xhci, > >>>>>>+ sizeof(xhci_plat_dwc3_xhci)); > >>>>>>+ 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"); > >>>>>I think you should just use device property: > >>>>> > >>>>This was suggested in an earlier series, but was rejected as it also added > >>>>the property as a device tree parameter. > >>>> > >>>>I think adding more device properties can be messy in the long run, especially if we > >>>>need to add them for many of the existing xhci quirks. > >>>>We also end up with a mix where some device properties are listed as device tree > >>>>parameters, and some not. > >>>> > >>>>Defining xhci quirks and platform data structure in headers shared with dwc3 and cdns3 > >>>>allow those drivers to easily set any existing xhci quirk, or other possible optional > >>>>callbacks. > >>>> > >>>>cdns3 driver is already doing this, but it includes the full xhci.h header. > >>>>This series cleans up that a bit so cdns3 will only include xhci quirk bits and > >>>>platform data structure. > >>>> > >>>>On the downside we add a couple xhci related header files to include/linux/usb/ > >>>>Let me know if you see any other issues I missed with this approach. > >>>The problem here is that these drivers are now coupled together, and > >>>that should not be taken lightly. We have a dependency hell in our > >>>hands with a lot of drivers, and the culprit is always platform data. > >>> > >>>Build-in device properties may be messy, but I would still say they > >>>are less messy than those quirk flags - you got to admit, they are a > >>>mess. The benefit from build-in properties is in any case the fact > >>>that they remove the need to couple these drivers together. > >>Agree, quirk bits are messy. Any suggestion that would work with > >>PCI xHCI devices, devicetree, and "pure" platform devices? > >I think xHCI driver should always be able to rely on being able to > >read this kind of information from the fwnode. If there is no actual > >firmware node (DT or ACPI), or if it's missing some information, the > >glue driver needs to populate software node for the xHCI. > > > >Right now I just want to avoid having to pass the quirks using > >platform data from drivers such as drivers/usb/cdns3/host.c and > >drivers/usb/dwc3/host.c to xHCI. > > > >One way we could do that is by defining compatibility ID for both of > >them that we provide using a single device property (like I guess DT > >does). Then based on that compatibility ID, xhci-plat.c can set the > >actual "static" quirk flags. That we could already do easily. How > >would that sound to you? > > This was my previous patch where I was using device tree property. Should we > go ahead with this approach? > > https://patchwork.kernel.org/project/linux-arm-msm/cover/1636353710-25582-1-git-send-email-quic_c_sanm@quicinc.com/ > > Any further changes to this ? > Actually the dT idea looks better than the platform data. Earlier, we decided to use if (of_device_is_compatible(dev->parent->of_node, "snps,dwc3")) check but I feel that we should just introduce a device tree param that specify this functionality (skipping phy init). we don't want xhci-plat to know that this is dwc3/xyz device. we expect it to be abstracted. Thanks, Pavan
On 29.3.2022 12.18, Sandeep Maheswaram (Temp) wrote: > Hi Mathias,Heikki > > On 3/25/2022 9:08 PM, Heikki Krogerus wrote: >> On Fri, Mar 25, 2022 at 04:33:27PM +0200, Mathias Nyman wrote: >>> On 25.3.2022 13.27, Heikki Krogerus wrote: >>>> On Fri, Mar 25, 2022 at 12:36:22AM +0200, Mathias Nyman wrote: >>>>> On 24.3.2022 14.27, Heikki Krogerus wrote: >>>>>> On Thu, Mar 24, 2022 at 12:07:11PM +0530, Sandeep Maheswaram wrote: >>>>>>> Currently the phy init is done from dwc3 and also xhci which makes the >>>>>>> runtime_usage value 2 for the phy which causes issue during runtime >>>>>>> suspend. When we run the below command the runtime_status still shows >>>>>>> active. >>>>>>> echo auto > /sys/bus/platform/devices/88e3000.phy/power/control >>>>>>> >>>>>>> dwc3 manages PHY by own DRD driver, so skip the management by >>>>>>> HCD core by setting this quirk. >>>>>>> >>>>>>> Signed-off-by: Sandeep Maheswaram <quic_c_sanm@quicinc.com> >>>>>>> --- >>>>>>> drivers/usb/dwc3/host.c | 13 +++++++++++++ >>>>>>> 1 file changed, 13 insertions(+) >>>>>>> >>>>>>> diff --git a/drivers/usb/dwc3/host.c b/drivers/usb/dwc3/host.c >>>>>>> index eda8719..d4fcf06 100644 >>>>>>> --- a/drivers/usb/dwc3/host.c >>>>>>> +++ b/drivers/usb/dwc3/host.c >>>>>>> @@ -13,6 +13,12 @@ >>>>>>> #include <linux/platform_device.h> >>>>>>> #include "core.h" >>>>>>> +#include <linux/usb/xhci-plat.h> >>>>>>> +#include <linux/usb/xhci-quirks.h> >>>>>>> + >>>>>>> +static const struct xhci_plat_priv xhci_plat_dwc3_xhci = { >>>>>>> + .quirks = XHCI_SKIP_PHY_INIT, >>>>>>> +}; >>>>>>> static void dwc3_host_fill_xhci_irq_res(struct dwc3 *dwc, >>>>>>> int irq, char *name) >>>>>>> @@ -122,6 +128,13 @@ int dwc3_host_init(struct dwc3 *dwc) >>>>>>> } >>>>>>> } >>>>>>> + ret = platform_device_add_data(xhci, &xhci_plat_dwc3_xhci, >>>>>>> + sizeof(xhci_plat_dwc3_xhci)); >>>>>>> + 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"); >>>>>> I think you should just use device property: >>>>>> >>>>> This was suggested in an earlier series, but was rejected as it also added >>>>> the property as a device tree parameter. >>>>> >>>>> I think adding more device properties can be messy in the long run, especially if we >>>>> need to add them for many of the existing xhci quirks. >>>>> We also end up with a mix where some device properties are listed as device tree >>>>> parameters, and some not. >>>>> >>>>> Defining xhci quirks and platform data structure in headers shared with dwc3 and cdns3 >>>>> allow those drivers to easily set any existing xhci quirk, or other possible optional >>>>> callbacks. >>>>> >>>>> cdns3 driver is already doing this, but it includes the full xhci.h header. >>>>> This series cleans up that a bit so cdns3 will only include xhci quirk bits and >>>>> platform data structure. >>>>> >>>>> On the downside we add a couple xhci related header files to include/linux/usb/ >>>>> Let me know if you see any other issues I missed with this approach. >>>> The problem here is that these drivers are now coupled together, and >>>> that should not be taken lightly. We have a dependency hell in our >>>> hands with a lot of drivers, and the culprit is always platform data. >>>> >>>> Build-in device properties may be messy, but I would still say they >>>> are less messy than those quirk flags - you got to admit, they are a >>>> mess. The benefit from build-in properties is in any case the fact >>>> that they remove the need to couple these drivers together. >>> Agree, quirk bits are messy. Any suggestion that would work with >>> PCI xHCI devices, devicetree, and "pure" platform devices? >> I think xHCI driver should always be able to rely on being able to >> read this kind of information from the fwnode. If there is no actual >> firmware node (DT or ACPI), or if it's missing some information, the >> glue driver needs to populate software node for the xHCI. >> >> Right now I just want to avoid having to pass the quirks using >> platform data from drivers such as drivers/usb/cdns3/host.c and >> drivers/usb/dwc3/host.c to xHCI. >> >> One way we could do that is by defining compatibility ID for both of >> them that we provide using a single device property (like I guess DT >> does). Then based on that compatibility ID, xhci-plat.c can set the >> actual "static" quirk flags. That we could already do easily. How >> would that sound to you? Sounds good. > > This was my previous patch where I was using device tree property. Should we go ahead with this approach? > > https://patchwork.kernel.org/project/linux-arm-msm/cover/1636353710-25582-1-git-send-email-quic_c_sanm@quicinc.com/ > > Any further changes to this ? By dropping the DT part of that series we get a similar built-in device property solution as Heikki initially suggested. How about adding the compatibility ID device property that was just suggested? Then matching the Id in xhci-plat.c against a static table containing Ids and xhci_plat_priv structures, with the needed quirks for dwc3. Thanks -Mathias
On Wed, Mar 30, 2022 at 08:47:34PM +0300, Mathias Nyman wrote: > On 29.3.2022 12.18, Sandeep Maheswaram (Temp) wrote: > > Hi Mathias,Heikki > > > > On 3/25/2022 9:08 PM, Heikki Krogerus wrote: > >> On Fri, Mar 25, 2022 at 04:33:27PM +0200, Mathias Nyman wrote: > >>> On 25.3.2022 13.27, Heikki Krogerus wrote: > >>>> On Fri, Mar 25, 2022 at 12:36:22AM +0200, Mathias Nyman wrote: > >>>>> On 24.3.2022 14.27, Heikki Krogerus wrote: > >>>>>> On Thu, Mar 24, 2022 at 12:07:11PM +0530, Sandeep Maheswaram wrote: > >>>>>>> Currently the phy init is done from dwc3 and also xhci which makes the > >>>>>>> runtime_usage value 2 for the phy which causes issue during runtime > >>>>>>> suspend. When we run the below command the runtime_status still shows > >>>>>>> active. > >>>>>>> echo auto > /sys/bus/platform/devices/88e3000.phy/power/control > >>>>>>> > >>>>>>> dwc3 manages PHY by own DRD driver, so skip the management by > >>>>>>> HCD core by setting this quirk. > >>>>>>> > >>>>>>> Signed-off-by: Sandeep Maheswaram <quic_c_sanm@quicinc.com> > >>>>>>> --- > >>>>>>> drivers/usb/dwc3/host.c | 13 +++++++++++++ > >>>>>>> 1 file changed, 13 insertions(+) > >>>>>>> > >>>>>>> diff --git a/drivers/usb/dwc3/host.c b/drivers/usb/dwc3/host.c > >>>>>>> index eda8719..d4fcf06 100644 > >>>>>>> --- a/drivers/usb/dwc3/host.c > >>>>>>> +++ b/drivers/usb/dwc3/host.c > >>>>>>> @@ -13,6 +13,12 @@ > >>>>>>> #include <linux/platform_device.h> > >>>>>>> #include "core.h" > >>>>>>> +#include <linux/usb/xhci-plat.h> > >>>>>>> +#include <linux/usb/xhci-quirks.h> > >>>>>>> + > >>>>>>> +static const struct xhci_plat_priv xhci_plat_dwc3_xhci = { > >>>>>>> + .quirks = XHCI_SKIP_PHY_INIT, > >>>>>>> +}; > >>>>>>> static void dwc3_host_fill_xhci_irq_res(struct dwc3 *dwc, > >>>>>>> int irq, char *name) > >>>>>>> @@ -122,6 +128,13 @@ int dwc3_host_init(struct dwc3 *dwc) > >>>>>>> } > >>>>>>> } > >>>>>>> + ret = platform_device_add_data(xhci, &xhci_plat_dwc3_xhci, > >>>>>>> + sizeof(xhci_plat_dwc3_xhci)); > >>>>>>> + 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"); > >>>>>> I think you should just use device property: > >>>>>> > >>>>> This was suggested in an earlier series, but was rejected as it also added > >>>>> the property as a device tree parameter. > >>>>> > >>>>> I think adding more device properties can be messy in the long run, especially if we > >>>>> need to add them for many of the existing xhci quirks. > >>>>> We also end up with a mix where some device properties are listed as device tree > >>>>> parameters, and some not. > >>>>> > >>>>> Defining xhci quirks and platform data structure in headers shared with dwc3 and cdns3 > >>>>> allow those drivers to easily set any existing xhci quirk, or other possible optional > >>>>> callbacks. > >>>>> > >>>>> cdns3 driver is already doing this, but it includes the full xhci.h header. > >>>>> This series cleans up that a bit so cdns3 will only include xhci quirk bits and > >>>>> platform data structure. > >>>>> > >>>>> On the downside we add a couple xhci related header files to include/linux/usb/ > >>>>> Let me know if you see any other issues I missed with this approach. > >>>> The problem here is that these drivers are now coupled together, and > >>>> that should not be taken lightly. We have a dependency hell in our > >>>> hands with a lot of drivers, and the culprit is always platform data. > >>>> > >>>> Build-in device properties may be messy, but I would still say they > >>>> are less messy than those quirk flags - you got to admit, they are a > >>>> mess. The benefit from build-in properties is in any case the fact > >>>> that they remove the need to couple these drivers together. > >>> Agree, quirk bits are messy. Any suggestion that would work with > >>> PCI xHCI devices, devicetree, and "pure" platform devices? > >> I think xHCI driver should always be able to rely on being able to > >> read this kind of information from the fwnode. If there is no actual > >> firmware node (DT or ACPI), or if it's missing some information, the > >> glue driver needs to populate software node for the xHCI. > >> > >> Right now I just want to avoid having to pass the quirks using > >> platform data from drivers such as drivers/usb/cdns3/host.c and > >> drivers/usb/dwc3/host.c to xHCI. > >> > >> One way we could do that is by defining compatibility ID for both of > >> them that we provide using a single device property (like I guess DT > >> does). Then based on that compatibility ID, xhci-plat.c can set the > >> actual "static" quirk flags. That we could already do easily. How > >> would that sound to you? > > Sounds good. > > > > > This was my previous patch where I was using device tree property. Should we go ahead with this approach? > > > > https://patchwork.kernel.org/project/linux-arm-msm/cover/1636353710-25582-1-git-send-email-quic_c_sanm@quicinc.com/ > > > > Any further changes to this ? > > By dropping the DT part of that series we get a similar built-in device property > solution as Heikki initially suggested. > > How about adding the compatibility ID device property that was just suggested? > Then matching the Id in xhci-plat.c against a static table containing Ids and > xhci_plat_priv structures, with the needed quirks for dwc3. There was a comment from Pavan. Is it still possible to get this detail from DT? I guess that would still be ideal, right? I have another question. Can't we now just assume that if the sysdev is the parent (or grandparent), then the phy initialization should always be skipped? In that case we could just do something like this: diff --git a/drivers/usb/host/xhci-plat.c b/drivers/usb/host/xhci-plat.c index 649ffd861b44e..1018b33488046 100644 --- a/drivers/usb/host/xhci-plat.c +++ b/drivers/usb/host/xhci-plat.c @@ -212,8 +212,12 @@ static int xhci_plat_probe(struct platform_device *pdev) #endif } - if (!sysdev) + if (sysdev) { + if (sysdev != &pdev->dev) + hcd->skip_phy_initialization = 1; + } else { sysdev = &pdev->dev; + } if (WARN_ON(!sysdev->dma_mask)) /* Platform did not initialize dma_mask */ I did not go through all the drivers that carefully, so I may have missed something, but it looks like the only drivers that can have the sysdev as the parent or grandparent are cdns3 and dwc3. thanks,
Hi Heikki, On Thu, Mar 31, 2022 at 02:16:53PM +0300, Heikki Krogerus wrote: > On Wed, Mar 30, 2022 at 08:47:34PM +0300, Mathias Nyman wrote: > > On 29.3.2022 12.18, Sandeep Maheswaram (Temp) wrote: > > > Hi Mathias,Heikki > > > > > > On 3/25/2022 9:08 PM, Heikki Krogerus wrote: > > >> On Fri, Mar 25, 2022 at 04:33:27PM +0200, Mathias Nyman wrote: > > >>> On 25.3.2022 13.27, Heikki Krogerus wrote: > > >>>> On Fri, Mar 25, 2022 at 12:36:22AM +0200, Mathias Nyman wrote: > > >>>>> On 24.3.2022 14.27, Heikki Krogerus wrote: > > >>>>>> On Thu, Mar 24, 2022 at 12:07:11PM +0530, Sandeep Maheswaram wrote: > > >>>>>>> Currently the phy init is done from dwc3 and also xhci which makes the > > >>>>>>> runtime_usage value 2 for the phy which causes issue during runtime > > >>>>>>> suspend. When we run the below command the runtime_status still shows > > >>>>>>> active. > > >>>>>>> echo auto > /sys/bus/platform/devices/88e3000.phy/power/control > > >>>>>>> > > >>>>>>> dwc3 manages PHY by own DRD driver, so skip the management by > > >>>>>>> HCD core by setting this quirk. > > >>>>>>> > > >>>>>>> Signed-off-by: Sandeep Maheswaram <quic_c_sanm@quicinc.com> > > >>>>>>> --- > > >>>>>>> drivers/usb/dwc3/host.c | 13 +++++++++++++ > > >>>>>>> 1 file changed, 13 insertions(+) > > >>>>>>> > > >>>>>>> diff --git a/drivers/usb/dwc3/host.c b/drivers/usb/dwc3/host.c > > >>>>>>> index eda8719..d4fcf06 100644 > > >>>>>>> --- a/drivers/usb/dwc3/host.c > > >>>>>>> +++ b/drivers/usb/dwc3/host.c > > >>>>>>> @@ -13,6 +13,12 @@ > > >>>>>>> #include <linux/platform_device.h> > > >>>>>>> #include "core.h" > > >>>>>>> +#include <linux/usb/xhci-plat.h> > > >>>>>>> +#include <linux/usb/xhci-quirks.h> > > >>>>>>> + > > >>>>>>> +static const struct xhci_plat_priv xhci_plat_dwc3_xhci = { > > >>>>>>> + .quirks = XHCI_SKIP_PHY_INIT, > > >>>>>>> +}; > > >>>>>>> static void dwc3_host_fill_xhci_irq_res(struct dwc3 *dwc, > > >>>>>>> int irq, char *name) > > >>>>>>> @@ -122,6 +128,13 @@ int dwc3_host_init(struct dwc3 *dwc) > > >>>>>>> } > > >>>>>>> } > > >>>>>>> + ret = platform_device_add_data(xhci, &xhci_plat_dwc3_xhci, > > >>>>>>> + sizeof(xhci_plat_dwc3_xhci)); > > >>>>>>> + 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"); > > >>>>>> I think you should just use device property: > > >>>>>> > > >>>>> This was suggested in an earlier series, but was rejected as it also added > > >>>>> the property as a device tree parameter. > > >>>>> > > >>>>> I think adding more device properties can be messy in the long run, especially if we > > >>>>> need to add them for many of the existing xhci quirks. > > >>>>> We also end up with a mix where some device properties are listed as device tree > > >>>>> parameters, and some not. > > >>>>> > > >>>>> Defining xhci quirks and platform data structure in headers shared with dwc3 and cdns3 > > >>>>> allow those drivers to easily set any existing xhci quirk, or other possible optional > > >>>>> callbacks. > > >>>>> > > >>>>> cdns3 driver is already doing this, but it includes the full xhci.h header. > > >>>>> This series cleans up that a bit so cdns3 will only include xhci quirk bits and > > >>>>> platform data structure. > > >>>>> > > >>>>> On the downside we add a couple xhci related header files to include/linux/usb/ > > >>>>> Let me know if you see any other issues I missed with this approach. > > >>>> The problem here is that these drivers are now coupled together, and > > >>>> that should not be taken lightly. We have a dependency hell in our > > >>>> hands with a lot of drivers, and the culprit is always platform data. > > >>>> > > >>>> Build-in device properties may be messy, but I would still say they > > >>>> are less messy than those quirk flags - you got to admit, they are a > > >>>> mess. The benefit from build-in properties is in any case the fact > > >>>> that they remove the need to couple these drivers together. > > >>> Agree, quirk bits are messy. Any suggestion that would work with > > >>> PCI xHCI devices, devicetree, and "pure" platform devices? > > >> I think xHCI driver should always be able to rely on being able to > > >> read this kind of information from the fwnode. If there is no actual > > >> firmware node (DT or ACPI), or if it's missing some information, the > > >> glue driver needs to populate software node for the xHCI. > > >> > > >> Right now I just want to avoid having to pass the quirks using > > >> platform data from drivers such as drivers/usb/cdns3/host.c and > > >> drivers/usb/dwc3/host.c to xHCI. > > >> > > >> One way we could do that is by defining compatibility ID for both of > > >> them that we provide using a single device property (like I guess DT > > >> does). Then based on that compatibility ID, xhci-plat.c can set the > > >> actual "static" quirk flags. That we could already do easily. How > > >> would that sound to you? > > > > Sounds good. > > > > > > > > This was my previous patch where I was using device tree property. Should we go ahead with this approach? > > > > > > https://patchwork.kernel.org/project/linux-arm-msm/cover/1636353710-25582-1-git-send-email-quic_c_sanm@quicinc.com/ > > > > > > Any further changes to this ? > > > > By dropping the DT part of that series we get a similar built-in device property > > solution as Heikki initially suggested. > > > > How about adding the compatibility ID device property that was just suggested? > > Then matching the Id in xhci-plat.c against a static table containing Ids and > > xhci_plat_priv structures, with the needed quirks for dwc3. > > There was a comment from Pavan. Is it still possible to get this > detail from DT? > I guess that would still be ideal, right? > I was suggesting if we can have device tree param like the patch sandeep pointed out. How would adding a compatible index to usb_xhci_of_match[] would work actually? I ask this because, dwc3/host.c creates platform device and it is not associated with any of_node, so of_driver_match_device() called from platform bus match method does not work. one way to achieve this would be by matching against sysdev. Something like below. Is it acceptible? diff --git a/drivers/usb/host/xhci-plat.c b/drivers/usb/host/xhci-plat.c index 649ffd8..bd5d055 100644 --- a/drivers/usb/host/xhci-plat.c +++ b/drivers/usb/host/xhci-plat.c @@ -126,6 +126,10 @@ static const struct xhci_plat_priv xhci_plat_brcm = { .quirks = XHCI_RESET_ON_RESUME, }; +static const struct xhci_plat_priv xhci_plat_dwc3 = { + .quirks = XHCI_SKIP_PHY_INIT, +}; + static const struct of_device_id usb_xhci_of_match[] = { { .compatible = "generic-xhci", @@ -167,6 +171,9 @@ static const struct of_device_id usb_xhci_of_match[] = { }, { .compatible = "brcm,bcm7445-xhci", .data = &xhci_plat_brcm, + }, { + .compatible = "snps,dwc3", + .data = &xhci_plat_dwc3, }, {}, }; @@ -274,6 +281,15 @@ static int xhci_plat_probe(struct platform_device *pdev) else priv_match = dev_get_platdata(&pdev->dev); + /* allow private data mapping with the sysdev compatible */ + if (!priv_match) { + struct of_device_id *match; + + match = of_match_device(usb_xhci_of_match, sysdev); + if (match) + priv_match = match->data; + } + if (priv_match) { priv = hcd_to_xhci_priv(hcd); /* Just copy data for now */ > I have another question. Can't we now just assume that if the sysdev > is the parent (or grandparent), then the phy initialization should > always be skipped? In that case we could just do something like this: > > diff --git a/drivers/usb/host/xhci-plat.c b/drivers/usb/host/xhci-plat.c > index 649ffd861b44e..1018b33488046 100644 > --- a/drivers/usb/host/xhci-plat.c > +++ b/drivers/usb/host/xhci-plat.c > @@ -212,8 +212,12 @@ static int xhci_plat_probe(struct platform_device *pdev) > #endif > } > > - if (!sysdev) > + if (sysdev) { > + if (sysdev != &pdev->dev) > + hcd->skip_phy_initialization = 1; > + } else { > sysdev = &pdev->dev; > + } > > if (WARN_ON(!sysdev->dma_mask)) > /* Platform did not initialize dma_mask */ > > > I did not go through all the drivers that carefully, so I may have > missed something, but it looks like the only drivers that can have the > sysdev as the parent or grandparent are cdns3 and dwc3. > I cross checked and these are two drivers that are creating xhci-plat device. So this patch would definitely work. However I am not sure in future if any device created via device tree would want to use this feature. For now, it looks good. It Mathias, Do you see any problem with this approach? Thanks, Pavan
Hi Heikki/Mathias, On Mon, Apr 04, 2022 at 01:55:16PM +0530, Pavan Kondeti wrote: > Hi Heikki, > > On Thu, Mar 31, 2022 at 02:16:53PM +0300, Heikki Krogerus wrote: > > On Wed, Mar 30, 2022 at 08:47:34PM +0300, Mathias Nyman wrote: > > > On 29.3.2022 12.18, Sandeep Maheswaram (Temp) wrote: > > > > Hi Mathias,Heikki > > > > > > > > On 3/25/2022 9:08 PM, Heikki Krogerus wrote: > > > >> On Fri, Mar 25, 2022 at 04:33:27PM +0200, Mathias Nyman wrote: > > > >>> On 25.3.2022 13.27, Heikki Krogerus wrote: > > > >>>> On Fri, Mar 25, 2022 at 12:36:22AM +0200, Mathias Nyman wrote: > > > >>>>> On 24.3.2022 14.27, Heikki Krogerus wrote: > > > >>>>>> On Thu, Mar 24, 2022 at 12:07:11PM +0530, Sandeep Maheswaram wrote: > > > >>>>>>> Currently the phy init is done from dwc3 and also xhci which makes the > > > >>>>>>> runtime_usage value 2 for the phy which causes issue during runtime > > > >>>>>>> suspend. When we run the below command the runtime_status still shows > > > >>>>>>> active. > > > >>>>>>> echo auto > /sys/bus/platform/devices/88e3000.phy/power/control > > > >>>>>>> > > > >>>>>>> dwc3 manages PHY by own DRD driver, so skip the management by > > > >>>>>>> HCD core by setting this quirk. > > > >>>>>>> > > > >>>>>>> Signed-off-by: Sandeep Maheswaram <quic_c_sanm@quicinc.com> > > > >>>>>>> --- > > > >>>>>>> drivers/usb/dwc3/host.c | 13 +++++++++++++ > > > >>>>>>> 1 file changed, 13 insertions(+) > > > >>>>>>> > > > >>>>>>> diff --git a/drivers/usb/dwc3/host.c b/drivers/usb/dwc3/host.c > > > >>>>>>> index eda8719..d4fcf06 100644 > > > >>>>>>> --- a/drivers/usb/dwc3/host.c > > > >>>>>>> +++ b/drivers/usb/dwc3/host.c > > > >>>>>>> @@ -13,6 +13,12 @@ > > > >>>>>>> #include <linux/platform_device.h> > > > >>>>>>> #include "core.h" > > > >>>>>>> +#include <linux/usb/xhci-plat.h> > > > >>>>>>> +#include <linux/usb/xhci-quirks.h> > > > >>>>>>> + > > > >>>>>>> +static const struct xhci_plat_priv xhci_plat_dwc3_xhci = { > > > >>>>>>> + .quirks = XHCI_SKIP_PHY_INIT, > > > >>>>>>> +}; > > > >>>>>>> static void dwc3_host_fill_xhci_irq_res(struct dwc3 *dwc, > > > >>>>>>> int irq, char *name) > > > >>>>>>> @@ -122,6 +128,13 @@ int dwc3_host_init(struct dwc3 *dwc) > > > >>>>>>> } > > > >>>>>>> } > > > >>>>>>> + ret = platform_device_add_data(xhci, &xhci_plat_dwc3_xhci, > > > >>>>>>> + sizeof(xhci_plat_dwc3_xhci)); > > > >>>>>>> + 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"); > > > >>>>>> I think you should just use device property: > > > >>>>>> > > > >>>>> This was suggested in an earlier series, but was rejected as it also added > > > >>>>> the property as a device tree parameter. > > > >>>>> > > > >>>>> I think adding more device properties can be messy in the long run, especially if we > > > >>>>> need to add them for many of the existing xhci quirks. > > > >>>>> We also end up with a mix where some device properties are listed as device tree > > > >>>>> parameters, and some not. > > > >>>>> > > > >>>>> Defining xhci quirks and platform data structure in headers shared with dwc3 and cdns3 > > > >>>>> allow those drivers to easily set any existing xhci quirk, or other possible optional > > > >>>>> callbacks. > > > >>>>> > > > >>>>> cdns3 driver is already doing this, but it includes the full xhci.h header. > > > >>>>> This series cleans up that a bit so cdns3 will only include xhci quirk bits and > > > >>>>> platform data structure. > > > >>>>> > > > >>>>> On the downside we add a couple xhci related header files to include/linux/usb/ > > > >>>>> Let me know if you see any other issues I missed with this approach. > > > >>>> The problem here is that these drivers are now coupled together, and > > > >>>> that should not be taken lightly. We have a dependency hell in our > > > >>>> hands with a lot of drivers, and the culprit is always platform data. > > > >>>> > > > >>>> Build-in device properties may be messy, but I would still say they > > > >>>> are less messy than those quirk flags - you got to admit, they are a > > > >>>> mess. The benefit from build-in properties is in any case the fact > > > >>>> that they remove the need to couple these drivers together. > > > >>> Agree, quirk bits are messy. Any suggestion that would work with > > > >>> PCI xHCI devices, devicetree, and "pure" platform devices? > > > >> I think xHCI driver should always be able to rely on being able to > > > >> read this kind of information from the fwnode. If there is no actual > > > >> firmware node (DT or ACPI), or if it's missing some information, the > > > >> glue driver needs to populate software node for the xHCI. > > > >> > > > >> Right now I just want to avoid having to pass the quirks using > > > >> platform data from drivers such as drivers/usb/cdns3/host.c and > > > >> drivers/usb/dwc3/host.c to xHCI. > > > >> > > > >> One way we could do that is by defining compatibility ID for both of > > > >> them that we provide using a single device property (like I guess DT > > > >> does). Then based on that compatibility ID, xhci-plat.c can set the > > > >> actual "static" quirk flags. That we could already do easily. How > > > >> would that sound to you? > > > > > > Sounds good. > > > > > > > > > > > This was my previous patch where I was using device tree property. Should we go ahead with this approach? > > > > > > > > https://patchwork.kernel.org/project/linux-arm-msm/cover/1636353710-25582-1-git-send-email-quic_c_sanm@quicinc.com/ > > > > > > > > Any further changes to this ? > > > > > > By dropping the DT part of that series we get a similar built-in device property > > > solution as Heikki initially suggested. > > > > > > How about adding the compatibility ID device property that was just suggested? > > > Then matching the Id in xhci-plat.c against a static table containing Ids and > > > xhci_plat_priv structures, with the needed quirks for dwc3. > > > > There was a comment from Pavan. Is it still possible to get this > > detail from DT? > > I guess that would still be ideal, right? > > > I was suggesting if we can have device tree param like the patch sandeep > pointed out. > > How would adding a compatible index to usb_xhci_of_match[] would work > actually? I ask this because, dwc3/host.c creates platform device and > it is not associated with any of_node, so of_driver_match_device() called > from platform bus match method does not work. one way to achieve this would > be by matching against sysdev. Something like below. Is it acceptible? > > diff --git a/drivers/usb/host/xhci-plat.c b/drivers/usb/host/xhci-plat.c > index 649ffd8..bd5d055 100644 > --- a/drivers/usb/host/xhci-plat.c > +++ b/drivers/usb/host/xhci-plat.c > @@ -126,6 +126,10 @@ static const struct xhci_plat_priv xhci_plat_brcm = { > .quirks = XHCI_RESET_ON_RESUME, > }; > > +static const struct xhci_plat_priv xhci_plat_dwc3 = { > + .quirks = XHCI_SKIP_PHY_INIT, > +}; > + > static const struct of_device_id usb_xhci_of_match[] = { > { > .compatible = "generic-xhci", > @@ -167,6 +171,9 @@ static const struct of_device_id usb_xhci_of_match[] = { > }, { > .compatible = "brcm,bcm7445-xhci", > .data = &xhci_plat_brcm, > + }, { > + .compatible = "snps,dwc3", > + .data = &xhci_plat_dwc3, > }, > {}, > }; > @@ -274,6 +281,15 @@ static int xhci_plat_probe(struct platform_device *pdev) > else > priv_match = dev_get_platdata(&pdev->dev); > > + /* allow private data mapping with the sysdev compatible */ > + if (!priv_match) { > + struct of_device_id *match; > + > + match = of_match_device(usb_xhci_of_match, sysdev); > + if (match) > + priv_match = match->data; > + } > + > if (priv_match) { > priv = hcd_to_xhci_priv(hcd); > /* Just copy data for now */ > > > I have another question. Can't we now just assume that if the sysdev > > is the parent (or grandparent), then the phy initialization should > > always be skipped? In that case we could just do something like this: > > > > diff --git a/drivers/usb/host/xhci-plat.c b/drivers/usb/host/xhci-plat.c > > index 649ffd861b44e..1018b33488046 100644 > > --- a/drivers/usb/host/xhci-plat.c > > +++ b/drivers/usb/host/xhci-plat.c > > @@ -212,8 +212,12 @@ static int xhci_plat_probe(struct platform_device *pdev) > > #endif > > } > > > > - if (!sysdev) > > + if (sysdev) { > > + if (sysdev != &pdev->dev) > > + hcd->skip_phy_initialization = 1; > > + } else { > > sysdev = &pdev->dev; > > + } > > > > if (WARN_ON(!sysdev->dma_mask)) > > /* Platform did not initialize dma_mask */ > > > > > > I did not go through all the drivers that carefully, so I may have > > missed something, but it looks like the only drivers that can have the > > sysdev as the parent or grandparent are cdns3 and dwc3. > > > I cross checked and these are two drivers that are creating xhci-plat device. > So this patch would definitely work. However I am not sure in future if any > device created via device tree would want to use this feature. For now, > it looks good. It Mathias, Do you see any problem with this approach? > Can you please provide your suggestions on this? We have discussed about 3 approaches here other than the whole platform data refactoring done. (1) Introduce a new dT property and expect dwc3/host.c to set this property to skip the phy initialization. (2) Allow platform private data match based on the sysdev. The diff I sent in the previous email. (3) Heikki's suggestion of relying on the fact that the users of phy skip init can be tested with sysdev != &pdev->dev check. Thanks, Pavan
On Wed, Apr 06, 2022 at 11:55:43AM +0530, Pavan Kondeti wrote: > Hi Heikki/Mathias, > > On Mon, Apr 04, 2022 at 01:55:16PM +0530, Pavan Kondeti wrote: > > Hi Heikki, > > > > On Thu, Mar 31, 2022 at 02:16:53PM +0300, Heikki Krogerus wrote: > > > On Wed, Mar 30, 2022 at 08:47:34PM +0300, Mathias Nyman wrote: > > > > On 29.3.2022 12.18, Sandeep Maheswaram (Temp) wrote: > > > > > Hi Mathias,Heikki > > > > > > > > > > On 3/25/2022 9:08 PM, Heikki Krogerus wrote: > > > > >> On Fri, Mar 25, 2022 at 04:33:27PM +0200, Mathias Nyman wrote: > > > > >>> On 25.3.2022 13.27, Heikki Krogerus wrote: > > > > >>>> On Fri, Mar 25, 2022 at 12:36:22AM +0200, Mathias Nyman wrote: > > > > >>>>> On 24.3.2022 14.27, Heikki Krogerus wrote: > > > > >>>>>> On Thu, Mar 24, 2022 at 12:07:11PM +0530, Sandeep Maheswaram wrote: > > > > >>>>>>> Currently the phy init is done from dwc3 and also xhci which makes the > > > > >>>>>>> runtime_usage value 2 for the phy which causes issue during runtime > > > > >>>>>>> suspend. When we run the below command the runtime_status still shows > > > > >>>>>>> active. > > > > >>>>>>> echo auto > /sys/bus/platform/devices/88e3000.phy/power/control > > > > >>>>>>> > > > > >>>>>>> dwc3 manages PHY by own DRD driver, so skip the management by > > > > >>>>>>> HCD core by setting this quirk. > > > > >>>>>>> > > > > >>>>>>> Signed-off-by: Sandeep Maheswaram <quic_c_sanm@quicinc.com> > > > > >>>>>>> --- > > > > >>>>>>> drivers/usb/dwc3/host.c | 13 +++++++++++++ > > > > >>>>>>> 1 file changed, 13 insertions(+) > > > > >>>>>>> > > > > >>>>>>> diff --git a/drivers/usb/dwc3/host.c b/drivers/usb/dwc3/host.c > > > > >>>>>>> index eda8719..d4fcf06 100644 > > > > >>>>>>> --- a/drivers/usb/dwc3/host.c > > > > >>>>>>> +++ b/drivers/usb/dwc3/host.c > > > > >>>>>>> @@ -13,6 +13,12 @@ > > > > >>>>>>> #include <linux/platform_device.h> > > > > >>>>>>> #include "core.h" > > > > >>>>>>> +#include <linux/usb/xhci-plat.h> > > > > >>>>>>> +#include <linux/usb/xhci-quirks.h> > > > > >>>>>>> + > > > > >>>>>>> +static const struct xhci_plat_priv xhci_plat_dwc3_xhci = { > > > > >>>>>>> + .quirks = XHCI_SKIP_PHY_INIT, > > > > >>>>>>> +}; > > > > >>>>>>> static void dwc3_host_fill_xhci_irq_res(struct dwc3 *dwc, > > > > >>>>>>> int irq, char *name) > > > > >>>>>>> @@ -122,6 +128,13 @@ int dwc3_host_init(struct dwc3 *dwc) > > > > >>>>>>> } > > > > >>>>>>> } > > > > >>>>>>> + ret = platform_device_add_data(xhci, &xhci_plat_dwc3_xhci, > > > > >>>>>>> + sizeof(xhci_plat_dwc3_xhci)); > > > > >>>>>>> + 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"); > > > > >>>>>> I think you should just use device property: > > > > >>>>>> > > > > >>>>> This was suggested in an earlier series, but was rejected as it also added > > > > >>>>> the property as a device tree parameter. > > > > >>>>> > > > > >>>>> I think adding more device properties can be messy in the long run, especially if we > > > > >>>>> need to add them for many of the existing xhci quirks. > > > > >>>>> We also end up with a mix where some device properties are listed as device tree > > > > >>>>> parameters, and some not. > > > > >>>>> > > > > >>>>> Defining xhci quirks and platform data structure in headers shared with dwc3 and cdns3 > > > > >>>>> allow those drivers to easily set any existing xhci quirk, or other possible optional > > > > >>>>> callbacks. > > > > >>>>> > > > > >>>>> cdns3 driver is already doing this, but it includes the full xhci.h header. > > > > >>>>> This series cleans up that a bit so cdns3 will only include xhci quirk bits and > > > > >>>>> platform data structure. > > > > >>>>> > > > > >>>>> On the downside we add a couple xhci related header files to include/linux/usb/ > > > > >>>>> Let me know if you see any other issues I missed with this approach. > > > > >>>> The problem here is that these drivers are now coupled together, and > > > > >>>> that should not be taken lightly. We have a dependency hell in our > > > > >>>> hands with a lot of drivers, and the culprit is always platform data. > > > > >>>> > > > > >>>> Build-in device properties may be messy, but I would still say they > > > > >>>> are less messy than those quirk flags - you got to admit, they are a > > > > >>>> mess. The benefit from build-in properties is in any case the fact > > > > >>>> that they remove the need to couple these drivers together. > > > > >>> Agree, quirk bits are messy. Any suggestion that would work with > > > > >>> PCI xHCI devices, devicetree, and "pure" platform devices? > > > > >> I think xHCI driver should always be able to rely on being able to > > > > >> read this kind of information from the fwnode. If there is no actual > > > > >> firmware node (DT or ACPI), or if it's missing some information, the > > > > >> glue driver needs to populate software node for the xHCI. > > > > >> > > > > >> Right now I just want to avoid having to pass the quirks using > > > > >> platform data from drivers such as drivers/usb/cdns3/host.c and > > > > >> drivers/usb/dwc3/host.c to xHCI. > > > > >> > > > > >> One way we could do that is by defining compatibility ID for both of > > > > >> them that we provide using a single device property (like I guess DT > > > > >> does). Then based on that compatibility ID, xhci-plat.c can set the > > > > >> actual "static" quirk flags. That we could already do easily. How > > > > >> would that sound to you? > > > > > > > > Sounds good. > > > > > > > > > > > > > > This was my previous patch where I was using device tree property. Should we go ahead with this approach? > > > > > > > > > > https://patchwork.kernel.org/project/linux-arm-msm/cover/1636353710-25582-1-git-send-email-quic_c_sanm@quicinc.com/ > > > > > > > > > > Any further changes to this ? > > > > > > > > By dropping the DT part of that series we get a similar built-in device property > > > > solution as Heikki initially suggested. > > > > > > > > How about adding the compatibility ID device property that was just suggested? > > > > Then matching the Id in xhci-plat.c against a static table containing Ids and > > > > xhci_plat_priv structures, with the needed quirks for dwc3. > > > > > > There was a comment from Pavan. Is it still possible to get this > > > detail from DT? > > > I guess that would still be ideal, right? > > > > > I was suggesting if we can have device tree param like the patch sandeep > > pointed out. > > > > How would adding a compatible index to usb_xhci_of_match[] would work > > actually? I ask this because, dwc3/host.c creates platform device and > > it is not associated with any of_node, so of_driver_match_device() called > > from platform bus match method does not work. one way to achieve this would > > be by matching against sysdev. Something like below. Is it acceptible? > > > > diff --git a/drivers/usb/host/xhci-plat.c b/drivers/usb/host/xhci-plat.c > > index 649ffd8..bd5d055 100644 > > --- a/drivers/usb/host/xhci-plat.c > > +++ b/drivers/usb/host/xhci-plat.c > > @@ -126,6 +126,10 @@ static const struct xhci_plat_priv xhci_plat_brcm = { > > .quirks = XHCI_RESET_ON_RESUME, > > }; > > > > +static const struct xhci_plat_priv xhci_plat_dwc3 = { > > + .quirks = XHCI_SKIP_PHY_INIT, > > +}; > > + > > static const struct of_device_id usb_xhci_of_match[] = { > > { > > .compatible = "generic-xhci", > > @@ -167,6 +171,9 @@ static const struct of_device_id usb_xhci_of_match[] = { > > }, { > > .compatible = "brcm,bcm7445-xhci", > > .data = &xhci_plat_brcm, > > + }, { > > + .compatible = "snps,dwc3", > > + .data = &xhci_plat_dwc3, > > }, > > {}, > > }; > > @@ -274,6 +281,15 @@ static int xhci_plat_probe(struct platform_device *pdev) > > else > > priv_match = dev_get_platdata(&pdev->dev); > > > > + /* allow private data mapping with the sysdev compatible */ > > + if (!priv_match) { > > + struct of_device_id *match; > > + > > + match = of_match_device(usb_xhci_of_match, sysdev); > > + if (match) > > + priv_match = match->data; > > + } > > + > > if (priv_match) { > > priv = hcd_to_xhci_priv(hcd); > > /* Just copy data for now */ > > > > > I have another question. Can't we now just assume that if the sysdev > > > is the parent (or grandparent), then the phy initialization should > > > always be skipped? In that case we could just do something like this: > > > > > > diff --git a/drivers/usb/host/xhci-plat.c b/drivers/usb/host/xhci-plat.c > > > index 649ffd861b44e..1018b33488046 100644 > > > --- a/drivers/usb/host/xhci-plat.c > > > +++ b/drivers/usb/host/xhci-plat.c > > > @@ -212,8 +212,12 @@ static int xhci_plat_probe(struct platform_device *pdev) > > > #endif > > > } > > > > > > - if (!sysdev) > > > + if (sysdev) { > > > + if (sysdev != &pdev->dev) > > > + hcd->skip_phy_initialization = 1; > > > + } else { > > > sysdev = &pdev->dev; > > > + } > > > > > > if (WARN_ON(!sysdev->dma_mask)) > > > /* Platform did not initialize dma_mask */ > > > > > > > > > I did not go through all the drivers that carefully, so I may have > > > missed something, but it looks like the only drivers that can have the > > > sysdev as the parent or grandparent are cdns3 and dwc3. > > > > > I cross checked and these are two drivers that are creating xhci-plat device. > > So this patch would definitely work. However I am not sure in future if any > > device created via device tree would want to use this feature. For now, > > it looks good. It Mathias, Do you see any problem with this approach? > > > > Can you please provide your suggestions on this? We have discussed about > 3 approaches here other than the whole platform data refactoring done. > > (1) Introduce a new dT property and expect dwc3/host.c to set this property > to skip the phy initialization. > (2) Allow platform private data match based on the sysdev. The diff I sent > in the previous email. > (3) Heikki's suggestion of relying on the fact that the users of phy skip > init can be tested with sysdev != &pdev->dev check. My vote is for option 3. That would allow us to sort this out properly later. I will in any case look at these drivers when I send the software node PM operations patch series. thanks,
Hi Sorry about the delayed response. On 6.4.2022 9.25, Pavan Kondeti wrote: > Hi Heikki/Mathias, > > On Mon, Apr 04, 2022 at 01:55:16PM +0530, Pavan Kondeti wrote: >> Hi Heikki, >> >> On Thu, Mar 31, 2022 at 02:16:53PM +0300, Heikki Krogerus wrote: >>> On Wed, Mar 30, 2022 at 08:47:34PM +0300, Mathias Nyman wrote: >>>> On 29.3.2022 12.18, Sandeep Maheswaram (Temp) wrote: >>>>> Hi Mathias,Heikki >>>>> >>>>> On 3/25/2022 9:08 PM, Heikki Krogerus wrote: >>>>>> On Fri, Mar 25, 2022 at 04:33:27PM +0200, Mathias Nyman wrote: >>>>>>> On 25.3.2022 13.27, Heikki Krogerus wrote: >>>>>>>> On Fri, Mar 25, 2022 at 12:36:22AM +0200, Mathias Nyman wrote: >>>>>>>>> On 24.3.2022 14.27, Heikki Krogerus wrote: >>>>>>>>>> On Thu, Mar 24, 2022 at 12:07:11PM +0530, Sandeep Maheswaram wrote: >>>>>>>>>>> Currently the phy init is done from dwc3 and also xhci which makes the >>>>>>>>>>> runtime_usage value 2 for the phy which causes issue during runtime >>>>>>>>>>> suspend. When we run the below command the runtime_status still shows >>>>>>>>>>> active. >>>>>>>>>>> echo auto > /sys/bus/platform/devices/88e3000.phy/power/control >>>>>>>>>>> >>>>>>>>>>> dwc3 manages PHY by own DRD driver, so skip the management by >>>>>>>>>>> HCD core by setting this quirk. >>>>>>>>>>> >>>>>>>>>>> Signed-off-by: Sandeep Maheswaram <quic_c_sanm@quicinc.com> >>>>>>>>>>> --- >>>>>>>>>>> drivers/usb/dwc3/host.c | 13 +++++++++++++ >>>>>>>>>>> 1 file changed, 13 insertions(+) >>>>>>>>>>> >>>>>>>>>>> diff --git a/drivers/usb/dwc3/host.c b/drivers/usb/dwc3/host.c >>>>>>>>>>> index eda8719..d4fcf06 100644 >>>>>>>>>>> --- a/drivers/usb/dwc3/host.c >>>>>>>>>>> +++ b/drivers/usb/dwc3/host.c >>>>>>>>>>> @@ -13,6 +13,12 @@ >>>>>>>>>>> #include <linux/platform_device.h> >>>>>>>>>>> #include "core.h" >>>>>>>>>>> +#include <linux/usb/xhci-plat.h> >>>>>>>>>>> +#include <linux/usb/xhci-quirks.h> >>>>>>>>>>> + >>>>>>>>>>> +static const struct xhci_plat_priv xhci_plat_dwc3_xhci = { >>>>>>>>>>> + .quirks = XHCI_SKIP_PHY_INIT, >>>>>>>>>>> +}; >>>>>>>>>>> static void dwc3_host_fill_xhci_irq_res(struct dwc3 *dwc, >>>>>>>>>>> int irq, char *name) >>>>>>>>>>> @@ -122,6 +128,13 @@ int dwc3_host_init(struct dwc3 *dwc) >>>>>>>>>>> } >>>>>>>>>>> } >>>>>>>>>>> + ret = platform_device_add_data(xhci, &xhci_plat_dwc3_xhci, >>>>>>>>>>> + sizeof(xhci_plat_dwc3_xhci)); >>>>>>>>>>> + 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"); >>>>>>>>>> I think you should just use device property: >>>>>>>>>> >>>>>>>>> This was suggested in an earlier series, but was rejected as it also added >>>>>>>>> the property as a device tree parameter. >>>>>>>>> >>>>>>>>> I think adding more device properties can be messy in the long run, especially if we >>>>>>>>> need to add them for many of the existing xhci quirks. >>>>>>>>> We also end up with a mix where some device properties are listed as device tree >>>>>>>>> parameters, and some not. >>>>>>>>> >>>>>>>>> Defining xhci quirks and platform data structure in headers shared with dwc3 and cdns3 >>>>>>>>> allow those drivers to easily set any existing xhci quirk, or other possible optional >>>>>>>>> callbacks. >>>>>>>>> >>>>>>>>> cdns3 driver is already doing this, but it includes the full xhci.h header. >>>>>>>>> This series cleans up that a bit so cdns3 will only include xhci quirk bits and >>>>>>>>> platform data structure. >>>>>>>>> >>>>>>>>> On the downside we add a couple xhci related header files to include/linux/usb/ >>>>>>>>> Let me know if you see any other issues I missed with this approach. >>>>>>>> The problem here is that these drivers are now coupled together, and >>>>>>>> that should not be taken lightly. We have a dependency hell in our >>>>>>>> hands with a lot of drivers, and the culprit is always platform data. >>>>>>>> >>>>>>>> Build-in device properties may be messy, but I would still say they >>>>>>>> are less messy than those quirk flags - you got to admit, they are a >>>>>>>> mess. The benefit from build-in properties is in any case the fact >>>>>>>> that they remove the need to couple these drivers together. >>>>>>> Agree, quirk bits are messy. Any suggestion that would work with >>>>>>> PCI xHCI devices, devicetree, and "pure" platform devices? >>>>>> I think xHCI driver should always be able to rely on being able to >>>>>> read this kind of information from the fwnode. If there is no actual >>>>>> firmware node (DT or ACPI), or if it's missing some information, the >>>>>> glue driver needs to populate software node for the xHCI. >>>>>> >>>>>> Right now I just want to avoid having to pass the quirks using >>>>>> platform data from drivers such as drivers/usb/cdns3/host.c and >>>>>> drivers/usb/dwc3/host.c to xHCI. >>>>>> >>>>>> One way we could do that is by defining compatibility ID for both of >>>>>> them that we provide using a single device property (like I guess DT >>>>>> does). Then based on that compatibility ID, xhci-plat.c can set the >>>>>> actual "static" quirk flags. That we could already do easily. How >>>>>> would that sound to you? >>>> >>>> Sounds good. >>>> >>>>> >>>>> This was my previous patch where I was using device tree property. Should we go ahead with this approach? >>>>> >>>>> https://patchwork.kernel.org/project/linux-arm-msm/cover/1636353710-25582-1-git-send-email-quic_c_sanm@quicinc.com/ >>>>> >>>>> Any further changes to this ? >>>> >>>> By dropping the DT part of that series we get a similar built-in device property >>>> solution as Heikki initially suggested. >>>> >>>> How about adding the compatibility ID device property that was just suggested? >>>> Then matching the Id in xhci-plat.c against a static table containing Ids and >>>> xhci_plat_priv structures, with the needed quirks for dwc3. >>> >>> There was a comment from Pavan. Is it still possible to get this >>> detail from DT? >>> I guess that would still be ideal, right? >>> >> I was suggesting if we can have device tree param like the patch sandeep >> pointed out. >> >> How would adding a compatible index to usb_xhci_of_match[] would work >> actually? I ask this because, dwc3/host.c creates platform device and >> it is not associated with any of_node, so of_driver_match_device() called >> from platform bus match method does not work. one way to achieve this would >> be by matching against sysdev. Something like below. Is it acceptible? >> >> diff --git a/drivers/usb/host/xhci-plat.c b/drivers/usb/host/xhci-plat.c >> index 649ffd8..bd5d055 100644 >> --- a/drivers/usb/host/xhci-plat.c >> +++ b/drivers/usb/host/xhci-plat.c >> @@ -126,6 +126,10 @@ static const struct xhci_plat_priv xhci_plat_brcm = { >> .quirks = XHCI_RESET_ON_RESUME, >> }; >> >> +static const struct xhci_plat_priv xhci_plat_dwc3 = { >> + .quirks = XHCI_SKIP_PHY_INIT, >> +}; >> + >> static const struct of_device_id usb_xhci_of_match[] = { >> { >> .compatible = "generic-xhci", >> @@ -167,6 +171,9 @@ static const struct of_device_id usb_xhci_of_match[] = { >> }, { >> .compatible = "brcm,bcm7445-xhci", >> .data = &xhci_plat_brcm, >> + }, { >> + .compatible = "snps,dwc3", >> + .data = &xhci_plat_dwc3, >> }, Isn't there a risk that xhci-plat now binds to the parent dwc3 device? competing with the similar of_match_table entry created in drivers/usb/dwc3/core.c >> {}, >> }; >> @@ -274,6 +281,15 @@ static int xhci_plat_probe(struct platform_device *pdev) >> else >> priv_match = dev_get_platdata(&pdev->dev); >> >> + /* allow private data mapping with the sysdev compatible */ >> + if (!priv_match) { >> + struct of_device_id *match; >> + >> + match = of_match_device(usb_xhci_of_match, sysdev); >> + if (match) >> + priv_match = match->data; >> + } >> + >> if (priv_match) { >> priv = hcd_to_xhci_priv(hcd); >> /* Just copy data for now */ >> >>> I have another question. Can't we now just assume that if the sysdev >>> is the parent (or grandparent), then the phy initialization should >>> always be skipped? In that case we could just do something like this: >>> >>> diff --git a/drivers/usb/host/xhci-plat.c b/drivers/usb/host/xhci-plat.c >>> index 649ffd861b44e..1018b33488046 100644 >>> --- a/drivers/usb/host/xhci-plat.c >>> +++ b/drivers/usb/host/xhci-plat.c >>> @@ -212,8 +212,12 @@ static int xhci_plat_probe(struct platform_device *pdev) >>> #endif >>> } >>> >>> - if (!sysdev) >>> + if (sysdev) { >>> + if (sysdev != &pdev->dev) >>> + hcd->skip_phy_initialization = 1; >>> + } else { >>> sysdev = &pdev->dev; >>> + } >>> >>> if (WARN_ON(!sysdev->dma_mask)) >>> /* Platform did not initialize dma_mask */ >>> >>> >>> I did not go through all the drivers that carefully, so I may have >>> missed something, but it looks like the only drivers that can have the >>> sysdev as the parent or grandparent are cdns3 and dwc3. >>> >> I cross checked and these are two drivers that are creating xhci-plat device. >> So this patch would definitely work. However I am not sure in future if any >> device created via device tree would want to use this feature. For now, >> it looks good. It Mathias, Do you see any problem with this approach? >> Would work for now but seems like a risk to assume this would hold for all future xhci platform devices. > > Can you please provide your suggestions on this? We have discussed about > 3 approaches here other than the whole platform data refactoring done. > > (1) Introduce a new dT property and expect dwc3/host.c to set this property > to skip the phy initialization. Adding one more device property to swnode in dwc3/host.c starts to look like the best option for now even if it didn't appeal initially. The place creating the xhci platform device should have best info on what properties are needed for the platform device. So this would be just like Heikki's first suggestion, or Sandeep's patches 2/3 and 3/3 in: https://patchwork.kernel.org/project/linux-arm-msm/cover/1636353710-25582-1-git-send-email-quic_c_sanm@quicinc.com/ but without the devicetree binding documentation part. > (2) Allow platform private data match based on the sysdev. The diff I sent > in the previous email -Possible risk of binding parent dwc3 to incorrect driver (xhci-plat), haven't looked at details. -Not working with ACPI dwc3 devices. Basic idea was similar to what I was thinking, but not matching with sysdev and misusing the of of_device_id table. Instead create a device property similar to "compatible". But looks like we would need to reinvent the wheel and create a some kind of swnode_device_id table, and custom match functions. So option 1 looks better to get this case fixed > (3) Heikki's suggestion of relying on the fact that the users of phy skip > init can be tested with sysdev != &pdev->dev check. Looks like a risky assumption for future, but could work as temporary solution So I think option 1 looks best, option 3 possible I also realize I don't know details of of_node, fwnode, swnode and device properties so all comments from Heikki and other who know more about this are welcome. Thanks -Mathias
Hi Mathias, On Wed, Apr 06, 2022 at 01:52:56PM +0300, Mathias Nyman wrote: > Hi > > Sorry about the delayed response. > > > On 6.4.2022 9.25, Pavan Kondeti wrote: > > Hi Heikki/Mathias, > > > > On Mon, Apr 04, 2022 at 01:55:16PM +0530, Pavan Kondeti wrote: > >> Hi Heikki, > >> > >> On Thu, Mar 31, 2022 at 02:16:53PM +0300, Heikki Krogerus wrote: > >>> On Wed, Mar 30, 2022 at 08:47:34PM +0300, Mathias Nyman wrote: > >>>> On 29.3.2022 12.18, Sandeep Maheswaram (Temp) wrote: > >>>>> Hi Mathias,Heikki > >>>>> > >>>>> On 3/25/2022 9:08 PM, Heikki Krogerus wrote: > >>>>>> On Fri, Mar 25, 2022 at 04:33:27PM +0200, Mathias Nyman wrote: > >>>>>>> On 25.3.2022 13.27, Heikki Krogerus wrote: > >>>>>>>> On Fri, Mar 25, 2022 at 12:36:22AM +0200, Mathias Nyman wrote: > >>>>>>>>> On 24.3.2022 14.27, Heikki Krogerus wrote: > >>>>>>>>>> On Thu, Mar 24, 2022 at 12:07:11PM +0530, Sandeep Maheswaram wrote: > >>>>>>>>>>> Currently the phy init is done from dwc3 and also xhci which makes the > >>>>>>>>>>> runtime_usage value 2 for the phy which causes issue during runtime > >>>>>>>>>>> suspend. When we run the below command the runtime_status still shows > >>>>>>>>>>> active. > >>>>>>>>>>> echo auto > /sys/bus/platform/devices/88e3000.phy/power/control > >>>>>>>>>>> > >>>>>>>>>>> dwc3 manages PHY by own DRD driver, so skip the management by > >>>>>>>>>>> HCD core by setting this quirk. > >>>>>>>>>>> > >>>>>>>>>>> Signed-off-by: Sandeep Maheswaram <quic_c_sanm@quicinc.com> > >>>>>>>>>>> --- > >>>>>>>>>>> drivers/usb/dwc3/host.c | 13 +++++++++++++ > >>>>>>>>>>> 1 file changed, 13 insertions(+) > >>>>>>>>>>> > >>>>>>>>>>> diff --git a/drivers/usb/dwc3/host.c b/drivers/usb/dwc3/host.c > >>>>>>>>>>> index eda8719..d4fcf06 100644 > >>>>>>>>>>> --- a/drivers/usb/dwc3/host.c > >>>>>>>>>>> +++ b/drivers/usb/dwc3/host.c > >>>>>>>>>>> @@ -13,6 +13,12 @@ > >>>>>>>>>>> #include <linux/platform_device.h> > >>>>>>>>>>> #include "core.h" > >>>>>>>>>>> +#include <linux/usb/xhci-plat.h> > >>>>>>>>>>> +#include <linux/usb/xhci-quirks.h> > >>>>>>>>>>> + > >>>>>>>>>>> +static const struct xhci_plat_priv xhci_plat_dwc3_xhci = { > >>>>>>>>>>> + .quirks = XHCI_SKIP_PHY_INIT, > >>>>>>>>>>> +}; > >>>>>>>>>>> static void dwc3_host_fill_xhci_irq_res(struct dwc3 *dwc, > >>>>>>>>>>> int irq, char *name) > >>>>>>>>>>> @@ -122,6 +128,13 @@ int dwc3_host_init(struct dwc3 *dwc) > >>>>>>>>>>> } > >>>>>>>>>>> } > >>>>>>>>>>> + ret = platform_device_add_data(xhci, &xhci_plat_dwc3_xhci, > >>>>>>>>>>> + sizeof(xhci_plat_dwc3_xhci)); > >>>>>>>>>>> + 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"); > >>>>>>>>>> I think you should just use device property: > >>>>>>>>>> > >>>>>>>>> This was suggested in an earlier series, but was rejected as it also added > >>>>>>>>> the property as a device tree parameter. > >>>>>>>>> > >>>>>>>>> I think adding more device properties can be messy in the long run, especially if we > >>>>>>>>> need to add them for many of the existing xhci quirks. > >>>>>>>>> We also end up with a mix where some device properties are listed as device tree > >>>>>>>>> parameters, and some not. > >>>>>>>>> > >>>>>>>>> Defining xhci quirks and platform data structure in headers shared with dwc3 and cdns3 > >>>>>>>>> allow those drivers to easily set any existing xhci quirk, or other possible optional > >>>>>>>>> callbacks. > >>>>>>>>> > >>>>>>>>> cdns3 driver is already doing this, but it includes the full xhci.h header. > >>>>>>>>> This series cleans up that a bit so cdns3 will only include xhci quirk bits and > >>>>>>>>> platform data structure. > >>>>>>>>> > >>>>>>>>> On the downside we add a couple xhci related header files to include/linux/usb/ > >>>>>>>>> Let me know if you see any other issues I missed with this approach. > >>>>>>>> The problem here is that these drivers are now coupled together, and > >>>>>>>> that should not be taken lightly. We have a dependency hell in our > >>>>>>>> hands with a lot of drivers, and the culprit is always platform data. > >>>>>>>> > >>>>>>>> Build-in device properties may be messy, but I would still say they > >>>>>>>> are less messy than those quirk flags - you got to admit, they are a > >>>>>>>> mess. The benefit from build-in properties is in any case the fact > >>>>>>>> that they remove the need to couple these drivers together. > >>>>>>> Agree, quirk bits are messy. Any suggestion that would work with > >>>>>>> PCI xHCI devices, devicetree, and "pure" platform devices? > >>>>>> I think xHCI driver should always be able to rely on being able to > >>>>>> read this kind of information from the fwnode. If there is no actual > >>>>>> firmware node (DT or ACPI), or if it's missing some information, the > >>>>>> glue driver needs to populate software node for the xHCI. > >>>>>> > >>>>>> Right now I just want to avoid having to pass the quirks using > >>>>>> platform data from drivers such as drivers/usb/cdns3/host.c and > >>>>>> drivers/usb/dwc3/host.c to xHCI. > >>>>>> > >>>>>> One way we could do that is by defining compatibility ID for both of > >>>>>> them that we provide using a single device property (like I guess DT > >>>>>> does). Then based on that compatibility ID, xhci-plat.c can set the > >>>>>> actual "static" quirk flags. That we could already do easily. How > >>>>>> would that sound to you? > >>>> > >>>> Sounds good. > >>>> > >>>>> > >>>>> This was my previous patch where I was using device tree property. Should we go ahead with this approach? > >>>>> > >>>>> https://patchwork.kernel.org/project/linux-arm-msm/cover/1636353710-25582-1-git-send-email-quic_c_sanm@quicinc.com/ > >>>>> > >>>>> Any further changes to this ? > >>>> > >>>> By dropping the DT part of that series we get a similar built-in device property > >>>> solution as Heikki initially suggested. > >>>> > >>>> How about adding the compatibility ID device property that was just suggested? > >>>> Then matching the Id in xhci-plat.c against a static table containing Ids and > >>>> xhci_plat_priv structures, with the needed quirks for dwc3. > >>> > >>> There was a comment from Pavan. Is it still possible to get this > >>> detail from DT? > >>> I guess that would still be ideal, right? > >>> > >> I was suggesting if we can have device tree param like the patch sandeep > >> pointed out. > >> > >> How would adding a compatible index to usb_xhci_of_match[] would work > >> actually? I ask this because, dwc3/host.c creates platform device and > >> it is not associated with any of_node, so of_driver_match_device() called > >> from platform bus match method does not work. one way to achieve this would > >> be by matching against sysdev. Something like below. Is it acceptible? > >> > >> diff --git a/drivers/usb/host/xhci-plat.c b/drivers/usb/host/xhci-plat.c > >> index 649ffd8..bd5d055 100644 > >> --- a/drivers/usb/host/xhci-plat.c > >> +++ b/drivers/usb/host/xhci-plat.c > >> @@ -126,6 +126,10 @@ static const struct xhci_plat_priv xhci_plat_brcm = { > >> .quirks = XHCI_RESET_ON_RESUME, > >> }; > >> > >> +static const struct xhci_plat_priv xhci_plat_dwc3 = { > >> + .quirks = XHCI_SKIP_PHY_INIT, > >> +}; > >> + > >> static const struct of_device_id usb_xhci_of_match[] = { > >> { > >> .compatible = "generic-xhci", > >> @@ -167,6 +171,9 @@ static const struct of_device_id usb_xhci_of_match[] = { > >> }, { > >> .compatible = "brcm,bcm7445-xhci", > >> .data = &xhci_plat_brcm, > >> + }, { > >> + .compatible = "snps,dwc3", > >> + .data = &xhci_plat_dwc3, > >> }, > > Isn't there a risk that xhci-plat now binds to the parent dwc3 device? > competing with the similar of_match_table entry created in drivers/usb/dwc3/core.c Sill of me. Yes, it does not work. Thanks for pointing it out. > > >> {}, > >> }; > >> @@ -274,6 +281,15 @@ static int xhci_plat_probe(struct platform_device *pdev) > >> else > >> priv_match = dev_get_platdata(&pdev->dev); > >> > >> + /* allow private data mapping with the sysdev compatible */ > >> + if (!priv_match) { > >> + struct of_device_id *match; > >> + > >> + match = of_match_device(usb_xhci_of_match, sysdev); > >> + if (match) > >> + priv_match = match->data; > >> + } > >> + > >> if (priv_match) { > >> priv = hcd_to_xhci_priv(hcd); > >> /* Just copy data for now */ > >> > >>> I have another question. Can't we now just assume that if the sysdev > >>> is the parent (or grandparent), then the phy initialization should > >>> always be skipped? In that case we could just do something like this: > >>> > >>> diff --git a/drivers/usb/host/xhci-plat.c b/drivers/usb/host/xhci-plat.c > >>> index 649ffd861b44e..1018b33488046 100644 > >>> --- a/drivers/usb/host/xhci-plat.c > >>> +++ b/drivers/usb/host/xhci-plat.c > >>> @@ -212,8 +212,12 @@ static int xhci_plat_probe(struct platform_device *pdev) > >>> #endif > >>> } > >>> > >>> - if (!sysdev) > >>> + if (sysdev) { > >>> + if (sysdev != &pdev->dev) > >>> + hcd->skip_phy_initialization = 1; > >>> + } else { > >>> sysdev = &pdev->dev; > >>> + } > >>> > >>> if (WARN_ON(!sysdev->dma_mask)) > >>> /* Platform did not initialize dma_mask */ > >>> > >>> > >>> I did not go through all the drivers that carefully, so I may have > >>> missed something, but it looks like the only drivers that can have the > >>> sysdev as the parent or grandparent are cdns3 and dwc3. > >>> > >> I cross checked and these are two drivers that are creating xhci-plat device. > >> So this patch would definitely work. However I am not sure in future if any > >> device created via device tree would want to use this feature. For now, > >> it looks good. It Mathias, Do you see any problem with this approach? > >> > > Would work for now but seems like a risk to assume this would hold for all future > xhci platform devices. > Agree that it may break in future for other drivers. > > > > Can you please provide your suggestions on this? We have discussed about > > 3 approaches here other than the whole platform data refactoring done. > > > > (1) Introduce a new dT property and expect dwc3/host.c to set this property > > to skip the phy initialization. > > Adding one more device property to swnode in dwc3/host.c starts > to look like the best option for now even if it didn't appeal initially. > > The place creating the xhci platform device should have best info on what properties > are needed for the platform device. > > So this would be just like Heikki's first suggestion, or Sandeep's patches 2/3 and 3/3 in: > https://patchwork.kernel.org/project/linux-arm-msm/cover/1636353710-25582-1-git-send-email-quic_c_sanm@quicinc.com/ > but without the devicetree binding documentation part. > Why do you say devicetree binding doc is not needed in this case? Possible that xhci-plat's device can be coming from dT and this param is passed to skip initialization of PHY. Thanks, Pavan
On 6.4.2022 14.01, Pavan Kondeti wrote: > Hi Mathias, > > On Wed, Apr 06, 2022 at 01:52:56PM +0300, Mathias Nyman wrote: >> Hi >> >> Sorry about the delayed response. >> >> >> On 6.4.2022 9.25, Pavan Kondeti wrote: >>> Hi Heikki/Mathias, >>> >>> On Mon, Apr 04, 2022 at 01:55:16PM +0530, Pavan Kondeti wrote: >>>> Hi Heikki, >>>> >>>> On Thu, Mar 31, 2022 at 02:16:53PM +0300, Heikki Krogerus wrote: >>>>> On Wed, Mar 30, 2022 at 08:47:34PM +0300, Mathias Nyman wrote: >>>>>> On 29.3.2022 12.18, Sandeep Maheswaram (Temp) wrote: >>>>>>> Hi Mathias,Heikki >>>>>>> >>>>>>> On 3/25/2022 9:08 PM, Heikki Krogerus wrote: >>>>>>>> On Fri, Mar 25, 2022 at 04:33:27PM +0200, Mathias Nyman wrote: >>>>>>>>> On 25.3.2022 13.27, Heikki Krogerus wrote: >>>>>>>>>> On Fri, Mar 25, 2022 at 12:36:22AM +0200, Mathias Nyman wrote: >>>>>>>>>>> On 24.3.2022 14.27, Heikki Krogerus wrote: >>>>>>>>>>>> On Thu, Mar 24, 2022 at 12:07:11PM +0530, Sandeep Maheswaram wrote: >>>>>>>>>>>>> Currently the phy init is done from dwc3 and also xhci which makes the >>>>>>>>>>>>> runtime_usage value 2 for the phy which causes issue during runtime >>>>>>>>>>>>> suspend. When we run the below command the runtime_status still shows >>>>>>>>>>>>> active. >>>>>>>>>>>>> echo auto > /sys/bus/platform/devices/88e3000.phy/power/control >>>>>>>>>>>>> >>>>>>>>>>>>> dwc3 manages PHY by own DRD driver, so skip the management by >>>>>>>>>>>>> HCD core by setting this quirk. >>>>>>>>>>>>> >>>>>>>>>>>>> Signed-off-by: Sandeep Maheswaram <quic_c_sanm@quicinc.com> >>>>>>>>>>>>> --- >>>>>>>>>>>>> drivers/usb/dwc3/host.c | 13 +++++++++++++ >>>>>>>>>>>>> 1 file changed, 13 insertions(+) >>>>>>>>>>>>> >>>>>>>>>>>>> diff --git a/drivers/usb/dwc3/host.c b/drivers/usb/dwc3/host.c >>>>>>>>>>>>> index eda8719..d4fcf06 100644 >>>>>>>>>>>>> --- a/drivers/usb/dwc3/host.c >>>>>>>>>>>>> +++ b/drivers/usb/dwc3/host.c >>>>>>>>>>>>> @@ -13,6 +13,12 @@ >>>>>>>>>>>>> #include <linux/platform_device.h> >>>>>>>>>>>>> #include "core.h" >>>>>>>>>>>>> +#include <linux/usb/xhci-plat.h> >>>>>>>>>>>>> +#include <linux/usb/xhci-quirks.h> >>>>>>>>>>>>> + >>>>>>>>>>>>> +static const struct xhci_plat_priv xhci_plat_dwc3_xhci = { >>>>>>>>>>>>> + .quirks = XHCI_SKIP_PHY_INIT, >>>>>>>>>>>>> +}; >>>>>>>>>>>>> static void dwc3_host_fill_xhci_irq_res(struct dwc3 *dwc, >>>>>>>>>>>>> int irq, char *name) >>>>>>>>>>>>> @@ -122,6 +128,13 @@ int dwc3_host_init(struct dwc3 *dwc) >>>>>>>>>>>>> } >>>>>>>>>>>>> } >>>>>>>>>>>>> + ret = platform_device_add_data(xhci, &xhci_plat_dwc3_xhci, >>>>>>>>>>>>> + sizeof(xhci_plat_dwc3_xhci)); >>>>>>>>>>>>> + 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"); >>>>>>>>>>>> I think you should just use device property: >>>>>>>>>>>> >>>>>>>>>>> This was suggested in an earlier series, but was rejected as it also added >>>>>>>>>>> the property as a device tree parameter. >>>>>>>>>>> >>>>>>>>>>> I think adding more device properties can be messy in the long run, especially if we >>>>>>>>>>> need to add them for many of the existing xhci quirks. >>>>>>>>>>> We also end up with a mix where some device properties are listed as device tree >>>>>>>>>>> parameters, and some not. >>>>>>>>>>> >>>>>>>>>>> Defining xhci quirks and platform data structure in headers shared with dwc3 and cdns3 >>>>>>>>>>> allow those drivers to easily set any existing xhci quirk, or other possible optional >>>>>>>>>>> callbacks. >>>>>>>>>>> >>>>>>>>>>> cdns3 driver is already doing this, but it includes the full xhci.h header. >>>>>>>>>>> This series cleans up that a bit so cdns3 will only include xhci quirk bits and >>>>>>>>>>> platform data structure. >>>>>>>>>>> >>>>>>>>>>> On the downside we add a couple xhci related header files to include/linux/usb/ >>>>>>>>>>> Let me know if you see any other issues I missed with this approach. >>>>>>>>>> The problem here is that these drivers are now coupled together, and >>>>>>>>>> that should not be taken lightly. We have a dependency hell in our >>>>>>>>>> hands with a lot of drivers, and the culprit is always platform data. >>>>>>>>>> >>>>>>>>>> Build-in device properties may be messy, but I would still say they >>>>>>>>>> are less messy than those quirk flags - you got to admit, they are a >>>>>>>>>> mess. The benefit from build-in properties is in any case the fact >>>>>>>>>> that they remove the need to couple these drivers together. >>>>>>>>> Agree, quirk bits are messy. Any suggestion that would work with >>>>>>>>> PCI xHCI devices, devicetree, and "pure" platform devices? >>>>>>>> I think xHCI driver should always be able to rely on being able to >>>>>>>> read this kind of information from the fwnode. If there is no actual >>>>>>>> firmware node (DT or ACPI), or if it's missing some information, the >>>>>>>> glue driver needs to populate software node for the xHCI. >>>>>>>> >>>>>>>> Right now I just want to avoid having to pass the quirks using >>>>>>>> platform data from drivers such as drivers/usb/cdns3/host.c and >>>>>>>> drivers/usb/dwc3/host.c to xHCI. >>>>>>>> >>>>>>>> One way we could do that is by defining compatibility ID for both of >>>>>>>> them that we provide using a single device property (like I guess DT >>>>>>>> does). Then based on that compatibility ID, xhci-plat.c can set the >>>>>>>> actual "static" quirk flags. That we could already do easily. How >>>>>>>> would that sound to you? >>>>>> >>>>>> Sounds good. >>>>>> >>>>>>> >>>>>>> This was my previous patch where I was using device tree property. Should we go ahead with this approach? >>>>>>> >>>>>>> https://patchwork.kernel.org/project/linux-arm-msm/cover/1636353710-25582-1-git-send-email-quic_c_sanm@quicinc.com/ >>>>>>> >>>>>>> Any further changes to this ? >>>>>> >>>>>> By dropping the DT part of that series we get a similar built-in device property >>>>>> solution as Heikki initially suggested. >>>>>> >>>>>> How about adding the compatibility ID device property that was just suggested? >>>>>> Then matching the Id in xhci-plat.c against a static table containing Ids and >>>>>> xhci_plat_priv structures, with the needed quirks for dwc3. >>>>> >>>>> There was a comment from Pavan. Is it still possible to get this >>>>> detail from DT? >>>>> I guess that would still be ideal, right? >>>>> >>>> I was suggesting if we can have device tree param like the patch sandeep >>>> pointed out. >>>> >>>> How would adding a compatible index to usb_xhci_of_match[] would work >>>> actually? I ask this because, dwc3/host.c creates platform device and >>>> it is not associated with any of_node, so of_driver_match_device() called >>>> from platform bus match method does not work. one way to achieve this would >>>> be by matching against sysdev. Something like below. Is it acceptible? >>>> >>>> diff --git a/drivers/usb/host/xhci-plat.c b/drivers/usb/host/xhci-plat.c >>>> index 649ffd8..bd5d055 100644 >>>> --- a/drivers/usb/host/xhci-plat.c >>>> +++ b/drivers/usb/host/xhci-plat.c >>>> @@ -126,6 +126,10 @@ static const struct xhci_plat_priv xhci_plat_brcm = { >>>> .quirks = XHCI_RESET_ON_RESUME, >>>> }; >>>> >>>> +static const struct xhci_plat_priv xhci_plat_dwc3 = { >>>> + .quirks = XHCI_SKIP_PHY_INIT, >>>> +}; >>>> + >>>> static const struct of_device_id usb_xhci_of_match[] = { >>>> { >>>> .compatible = "generic-xhci", >>>> @@ -167,6 +171,9 @@ static const struct of_device_id usb_xhci_of_match[] = { >>>> }, { >>>> .compatible = "brcm,bcm7445-xhci", >>>> .data = &xhci_plat_brcm, >>>> + }, { >>>> + .compatible = "snps,dwc3", >>>> + .data = &xhci_plat_dwc3, >>>> }, >> >> Isn't there a risk that xhci-plat now binds to the parent dwc3 device? >> competing with the similar of_match_table entry created in drivers/usb/dwc3/core.c > > Sill of me. Yes, it does not work. Thanks for pointing it out. > >> >>>> {}, >>>> }; >>>> @@ -274,6 +281,15 @@ static int xhci_plat_probe(struct platform_device *pdev) >>>> else >>>> priv_match = dev_get_platdata(&pdev->dev); >>>> >>>> + /* allow private data mapping with the sysdev compatible */ >>>> + if (!priv_match) { >>>> + struct of_device_id *match; >>>> + >>>> + match = of_match_device(usb_xhci_of_match, sysdev); >>>> + if (match) >>>> + priv_match = match->data; >>>> + } >>>> + >>>> if (priv_match) { >>>> priv = hcd_to_xhci_priv(hcd); >>>> /* Just copy data for now */ >>>> >>>>> I have another question. Can't we now just assume that if the sysdev >>>>> is the parent (or grandparent), then the phy initialization should >>>>> always be skipped? In that case we could just do something like this: >>>>> >>>>> diff --git a/drivers/usb/host/xhci-plat.c b/drivers/usb/host/xhci-plat.c >>>>> index 649ffd861b44e..1018b33488046 100644 >>>>> --- a/drivers/usb/host/xhci-plat.c >>>>> +++ b/drivers/usb/host/xhci-plat.c >>>>> @@ -212,8 +212,12 @@ static int xhci_plat_probe(struct platform_device *pdev) >>>>> #endif >>>>> } >>>>> >>>>> - if (!sysdev) >>>>> + if (sysdev) { >>>>> + if (sysdev != &pdev->dev) >>>>> + hcd->skip_phy_initialization = 1; >>>>> + } else { >>>>> sysdev = &pdev->dev; >>>>> + } >>>>> >>>>> if (WARN_ON(!sysdev->dma_mask)) >>>>> /* Platform did not initialize dma_mask */ >>>>> >>>>> >>>>> I did not go through all the drivers that carefully, so I may have >>>>> missed something, but it looks like the only drivers that can have the >>>>> sysdev as the parent or grandparent are cdns3 and dwc3. >>>>> >>>> I cross checked and these are two drivers that are creating xhci-plat device. >>>> So this patch would definitely work. However I am not sure in future if any >>>> device created via device tree would want to use this feature. For now, >>>> it looks good. It Mathias, Do you see any problem with this approach? >>>> >> >> Would work for now but seems like a risk to assume this would hold for all future >> xhci platform devices. >> > Agree that it may break in future for other drivers. > >>> >>> Can you please provide your suggestions on this? We have discussed about >>> 3 approaches here other than the whole platform data refactoring done. >>> >>> (1) Introduce a new dT property and expect dwc3/host.c to set this property >>> to skip the phy initialization. >> >> Adding one more device property to swnode in dwc3/host.c starts >> to look like the best option for now even if it didn't appeal initially. >> >> The place creating the xhci platform device should have best info on what properties >> are needed for the platform device. >> >> So this would be just like Heikki's first suggestion, or Sandeep's patches 2/3 and 3/3 in: >> https://patchwork.kernel.org/project/linux-arm-msm/cover/1636353710-25582-1-git-send-email-quic_c_sanm@quicinc.com/ >> but without the devicetree binding documentation part. >> > Why do you say devicetree binding doc is not needed in this case? Possible > that xhci-plat's device can be coming from dT and this param is passed to > skip initialization of PHY. Adding it is fine by me. It was rejected earlier by Rob Herring But we can live without it. In dwc3 and cdns3 "pure" platform case we add the device property when creating the device. In dT case we should have a compatible entry, and we can add a matching entry in usb_xhci_of_match[] which sets the needed quirk. Thanks -Mathias
Hi Mathias, On Wed, Apr 06, 2022 at 03:18:17PM +0300, Mathias Nyman wrote: > On 6.4.2022 14.01, Pavan Kondeti wrote: > > Hi Mathias, > > > > On Wed, Apr 06, 2022 at 01:52:56PM +0300, Mathias Nyman wrote: > >> Hi > >> > >> Sorry about the delayed response. > >> > >> > >> On 6.4.2022 9.25, Pavan Kondeti wrote: > >>> Hi Heikki/Mathias, > >>> > >>> On Mon, Apr 04, 2022 at 01:55:16PM +0530, Pavan Kondeti wrote: > >>>> Hi Heikki, > >>>> > >>>> On Thu, Mar 31, 2022 at 02:16:53PM +0300, Heikki Krogerus wrote: > >>>>> On Wed, Mar 30, 2022 at 08:47:34PM +0300, Mathias Nyman wrote: > >>>>>> On 29.3.2022 12.18, Sandeep Maheswaram (Temp) wrote: > >>>>>>> Hi Mathias,Heikki > >>>>>>> > >>>>>>> On 3/25/2022 9:08 PM, Heikki Krogerus wrote: > >>>>>>>> On Fri, Mar 25, 2022 at 04:33:27PM +0200, Mathias Nyman wrote: > >>>>>>>>> On 25.3.2022 13.27, Heikki Krogerus wrote: > >>>>>>>>>> On Fri, Mar 25, 2022 at 12:36:22AM +0200, Mathias Nyman wrote: > >>>>>>>>>>> On 24.3.2022 14.27, Heikki Krogerus wrote: > >>>>>>>>>>>> On Thu, Mar 24, 2022 at 12:07:11PM +0530, Sandeep Maheswaram wrote: > >>>>>>>>>>>>> Currently the phy init is done from dwc3 and also xhci which makes the > >>>>>>>>>>>>> runtime_usage value 2 for the phy which causes issue during runtime > >>>>>>>>>>>>> suspend. When we run the below command the runtime_status still shows > >>>>>>>>>>>>> active. > >>>>>>>>>>>>> echo auto > /sys/bus/platform/devices/88e3000.phy/power/control > >>>>>>>>>>>>> > >>>>>>>>>>>>> dwc3 manages PHY by own DRD driver, so skip the management by > >>>>>>>>>>>>> HCD core by setting this quirk. > >>>>>>>>>>>>> > >>>>>>>>>>>>> Signed-off-by: Sandeep Maheswaram <quic_c_sanm@quicinc.com> > >>>>>>>>>>>>> --- > >>>>>>>>>>>>> drivers/usb/dwc3/host.c | 13 +++++++++++++ > >>>>>>>>>>>>> 1 file changed, 13 insertions(+) > >>>>>>>>>>>>> > >>>>>>>>>>>>> diff --git a/drivers/usb/dwc3/host.c b/drivers/usb/dwc3/host.c > >>>>>>>>>>>>> index eda8719..d4fcf06 100644 > >>>>>>>>>>>>> --- a/drivers/usb/dwc3/host.c > >>>>>>>>>>>>> +++ b/drivers/usb/dwc3/host.c > >>>>>>>>>>>>> @@ -13,6 +13,12 @@ > >>>>>>>>>>>>> #include <linux/platform_device.h> > >>>>>>>>>>>>> #include "core.h" > >>>>>>>>>>>>> +#include <linux/usb/xhci-plat.h> > >>>>>>>>>>>>> +#include <linux/usb/xhci-quirks.h> > >>>>>>>>>>>>> + > >>>>>>>>>>>>> +static const struct xhci_plat_priv xhci_plat_dwc3_xhci = { > >>>>>>>>>>>>> + .quirks = XHCI_SKIP_PHY_INIT, > >>>>>>>>>>>>> +}; > >>>>>>>>>>>>> static void dwc3_host_fill_xhci_irq_res(struct dwc3 *dwc, > >>>>>>>>>>>>> int irq, char *name) > >>>>>>>>>>>>> @@ -122,6 +128,13 @@ int dwc3_host_init(struct dwc3 *dwc) > >>>>>>>>>>>>> } > >>>>>>>>>>>>> } > >>>>>>>>>>>>> + ret = platform_device_add_data(xhci, &xhci_plat_dwc3_xhci, > >>>>>>>>>>>>> + sizeof(xhci_plat_dwc3_xhci)); > >>>>>>>>>>>>> + 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"); > >>>>>>>>>>>> I think you should just use device property: > >>>>>>>>>>>> > >>>>>>>>>>> This was suggested in an earlier series, but was rejected as it also added > >>>>>>>>>>> the property as a device tree parameter. > >>>>>>>>>>> > >>>>>>>>>>> I think adding more device properties can be messy in the long run, especially if we > >>>>>>>>>>> need to add them for many of the existing xhci quirks. > >>>>>>>>>>> We also end up with a mix where some device properties are listed as device tree > >>>>>>>>>>> parameters, and some not. > >>>>>>>>>>> > >>>>>>>>>>> Defining xhci quirks and platform data structure in headers shared with dwc3 and cdns3 > >>>>>>>>>>> allow those drivers to easily set any existing xhci quirk, or other possible optional > >>>>>>>>>>> callbacks. > >>>>>>>>>>> > >>>>>>>>>>> cdns3 driver is already doing this, but it includes the full xhci.h header. > >>>>>>>>>>> This series cleans up that a bit so cdns3 will only include xhci quirk bits and > >>>>>>>>>>> platform data structure. > >>>>>>>>>>> > >>>>>>>>>>> On the downside we add a couple xhci related header files to include/linux/usb/ > >>>>>>>>>>> Let me know if you see any other issues I missed with this approach. > >>>>>>>>>> The problem here is that these drivers are now coupled together, and > >>>>>>>>>> that should not be taken lightly. We have a dependency hell in our > >>>>>>>>>> hands with a lot of drivers, and the culprit is always platform data. > >>>>>>>>>> > >>>>>>>>>> Build-in device properties may be messy, but I would still say they > >>>>>>>>>> are less messy than those quirk flags - you got to admit, they are a > >>>>>>>>>> mess. The benefit from build-in properties is in any case the fact > >>>>>>>>>> that they remove the need to couple these drivers together. > >>>>>>>>> Agree, quirk bits are messy. Any suggestion that would work with > >>>>>>>>> PCI xHCI devices, devicetree, and "pure" platform devices? > >>>>>>>> I think xHCI driver should always be able to rely on being able to > >>>>>>>> read this kind of information from the fwnode. If there is no actual > >>>>>>>> firmware node (DT or ACPI), or if it's missing some information, the > >>>>>>>> glue driver needs to populate software node for the xHCI. > >>>>>>>> > >>>>>>>> Right now I just want to avoid having to pass the quirks using > >>>>>>>> platform data from drivers such as drivers/usb/cdns3/host.c and > >>>>>>>> drivers/usb/dwc3/host.c to xHCI. > >>>>>>>> > >>>>>>>> One way we could do that is by defining compatibility ID for both of > >>>>>>>> them that we provide using a single device property (like I guess DT > >>>>>>>> does). Then based on that compatibility ID, xhci-plat.c can set the > >>>>>>>> actual "static" quirk flags. That we could already do easily. How > >>>>>>>> would that sound to you? > >>>>>> > >>>>>> Sounds good. > >>>>>> > >>>>>>> > >>>>>>> This was my previous patch where I was using device tree property. Should we go ahead with this approach? > >>>>>>> > >>>>>>> https://patchwork.kernel.org/project/linux-arm-msm/cover/1636353710-25582-1-git-send-email-quic_c_sanm@quicinc.com/ > >>>>>>> > >>>>>>> Any further changes to this ? > >>>>>> > >>>>>> By dropping the DT part of that series we get a similar built-in device property > >>>>>> solution as Heikki initially suggested. > >>>>>> > >>>>>> How about adding the compatibility ID device property that was just suggested? > >>>>>> Then matching the Id in xhci-plat.c against a static table containing Ids and > >>>>>> xhci_plat_priv structures, with the needed quirks for dwc3. > >>>>> > >>>>> There was a comment from Pavan. Is it still possible to get this > >>>>> detail from DT? > >>>>> I guess that would still be ideal, right? > >>>>> > >>>> I was suggesting if we can have device tree param like the patch sandeep > >>>> pointed out. > >>>> > >>>> How would adding a compatible index to usb_xhci_of_match[] would work > >>>> actually? I ask this because, dwc3/host.c creates platform device and > >>>> it is not associated with any of_node, so of_driver_match_device() called > >>>> from platform bus match method does not work. one way to achieve this would > >>>> be by matching against sysdev. Something like below. Is it acceptible? > >>>> > >>>> diff --git a/drivers/usb/host/xhci-plat.c b/drivers/usb/host/xhci-plat.c > >>>> index 649ffd8..bd5d055 100644 > >>>> --- a/drivers/usb/host/xhci-plat.c > >>>> +++ b/drivers/usb/host/xhci-plat.c > >>>> @@ -126,6 +126,10 @@ static const struct xhci_plat_priv xhci_plat_brcm = { > >>>> .quirks = XHCI_RESET_ON_RESUME, > >>>> }; > >>>> > >>>> +static const struct xhci_plat_priv xhci_plat_dwc3 = { > >>>> + .quirks = XHCI_SKIP_PHY_INIT, > >>>> +}; > >>>> + > >>>> static const struct of_device_id usb_xhci_of_match[] = { > >>>> { > >>>> .compatible = "generic-xhci", > >>>> @@ -167,6 +171,9 @@ static const struct of_device_id usb_xhci_of_match[] = { > >>>> }, { > >>>> .compatible = "brcm,bcm7445-xhci", > >>>> .data = &xhci_plat_brcm, > >>>> + }, { > >>>> + .compatible = "snps,dwc3", > >>>> + .data = &xhci_plat_dwc3, > >>>> }, > >> > >> Isn't there a risk that xhci-plat now binds to the parent dwc3 device? > >> competing with the similar of_match_table entry created in drivers/usb/dwc3/core.c > > > > Sill of me. Yes, it does not work. Thanks for pointing it out. > > > >> > >>>> {}, > >>>> }; > >>>> @@ -274,6 +281,15 @@ static int xhci_plat_probe(struct platform_device *pdev) > >>>> else > >>>> priv_match = dev_get_platdata(&pdev->dev); > >>>> > >>>> + /* allow private data mapping with the sysdev compatible */ > >>>> + if (!priv_match) { > >>>> + struct of_device_id *match; > >>>> + > >>>> + match = of_match_device(usb_xhci_of_match, sysdev); > >>>> + if (match) > >>>> + priv_match = match->data; > >>>> + } > >>>> + > >>>> if (priv_match) { > >>>> priv = hcd_to_xhci_priv(hcd); > >>>> /* Just copy data for now */ > >>>> > >>>>> I have another question. Can't we now just assume that if the sysdev > >>>>> is the parent (or grandparent), then the phy initialization should > >>>>> always be skipped? In that case we could just do something like this: > >>>>> > >>>>> diff --git a/drivers/usb/host/xhci-plat.c b/drivers/usb/host/xhci-plat.c > >>>>> index 649ffd861b44e..1018b33488046 100644 > >>>>> --- a/drivers/usb/host/xhci-plat.c > >>>>> +++ b/drivers/usb/host/xhci-plat.c > >>>>> @@ -212,8 +212,12 @@ static int xhci_plat_probe(struct platform_device *pdev) > >>>>> #endif > >>>>> } > >>>>> > >>>>> - if (!sysdev) > >>>>> + if (sysdev) { > >>>>> + if (sysdev != &pdev->dev) > >>>>> + hcd->skip_phy_initialization = 1; > >>>>> + } else { > >>>>> sysdev = &pdev->dev; > >>>>> + } > >>>>> > >>>>> if (WARN_ON(!sysdev->dma_mask)) > >>>>> /* Platform did not initialize dma_mask */ > >>>>> > >>>>> > >>>>> I did not go through all the drivers that carefully, so I may have > >>>>> missed something, but it looks like the only drivers that can have the > >>>>> sysdev as the parent or grandparent are cdns3 and dwc3. > >>>>> > >>>> I cross checked and these are two drivers that are creating xhci-plat device. > >>>> So this patch would definitely work. However I am not sure in future if any > >>>> device created via device tree would want to use this feature. For now, > >>>> it looks good. It Mathias, Do you see any problem with this approach? > >>>> > >> > >> Would work for now but seems like a risk to assume this would hold for all future > >> xhci platform devices. > >> > > Agree that it may break in future for other drivers. > > > >>> > >>> Can you please provide your suggestions on this? We have discussed about > >>> 3 approaches here other than the whole platform data refactoring done. > >>> > >>> (1) Introduce a new dT property and expect dwc3/host.c to set this property > >>> to skip the phy initialization. > >> > >> Adding one more device property to swnode in dwc3/host.c starts > >> to look like the best option for now even if it didn't appeal initially. > >> > >> The place creating the xhci platform device should have best info on what properties > >> are needed for the platform device. > >> > >> So this would be just like Heikki's first suggestion, or Sandeep's patches 2/3 and 3/3 in: > >> https://patchwork.kernel.org/project/linux-arm-msm/cover/1636353710-25582-1-git-send-email-quic_c_sanm@quicinc.com/ > >> but without the devicetree binding documentation part. > >> > > Why do you say devicetree binding doc is not needed in this case? Possible > > that xhci-plat's device can be coming from dT and this param is passed to > > skip initialization of PHY. > > Adding it is fine by me. It was rejected earlier by Rob Herring > > But we can live without it. > In dwc3 and cdns3 "pure" platform case we add the device property when > creating the device. > > In dT case we should have a compatible entry, and we can add a matching > entry in usb_xhci_of_match[] which sets the needed quirk. > Got it. Thanks for the explanation. We can go with Sandeep's patches 2/3 and 3/3 which add dT param and can be set by the dwc3/host.c Thanks, Pavan
diff --git a/drivers/usb/dwc3/host.c b/drivers/usb/dwc3/host.c index eda8719..d4fcf06 100644 --- a/drivers/usb/dwc3/host.c +++ b/drivers/usb/dwc3/host.c @@ -13,6 +13,12 @@ #include <linux/platform_device.h> #include "core.h" +#include <linux/usb/xhci-plat.h> +#include <linux/usb/xhci-quirks.h> + +static const struct xhci_plat_priv xhci_plat_dwc3_xhci = { + .quirks = XHCI_SKIP_PHY_INIT, +}; static void dwc3_host_fill_xhci_irq_res(struct dwc3 *dwc, int irq, char *name) @@ -122,6 +128,13 @@ int dwc3_host_init(struct dwc3 *dwc) } } + ret = platform_device_add_data(xhci, &xhci_plat_dwc3_xhci, + sizeof(xhci_plat_dwc3_xhci)); + 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");
Currently the phy init is done from dwc3 and also xhci which makes the runtime_usage value 2 for the phy which causes issue during runtime suspend. When we run the below command the runtime_status still shows active. echo auto > /sys/bus/platform/devices/88e3000.phy/power/control dwc3 manages PHY by own DRD driver, so skip the management by HCD core by setting this quirk. Signed-off-by: Sandeep Maheswaram <quic_c_sanm@quicinc.com> --- drivers/usb/dwc3/host.c | 13 +++++++++++++ 1 file changed, 13 insertions(+)