From: Kishon Vijay Abraham I <hidden> Date: 2021-08-26 11:14:40
Cold plugged USB device was not detected on certain PCIe USB cards
(like Inateck card connected to AM64 EVM or connected to J7200 EVM).
Re-plugging the USB device always gets it enumerated.
This issue was discussed in
https://lore.kernel.org/r/772e4001-178e-4918-032c-6e625bdded24@ti.com
and
https://bugzilla.kernel.org/show_bug.cgi?id=214021
So the suggested solution is to register both root hubs along with the
second hcd for xhci.
RFC Patch series can be found at [1]
v1 Patch series can be found at [3]
Changes from RFC:
1) Mathias identified potential issues with the RFC patch [2] and suggested
the solution to use HCD flags. This series implements it.
Changes from v1:
1) Fixed code comments pointed out by Alan Stern
2) Renamed the HCD flag variable to "HCD_FLAG_DEFER_RH_REGISTER" from
"HCD_FLAG_DEFER_PRI_RH_REGISTER"
[1] -> https://lore.kernel.org/linux-usb/20210824105302.25382-1-kishon@ti.com/
[2] -> https://lore.kernel.org/linux-usb/06693934-28f2-d59e-b004-62cabd3f9e8e@linux.intel.com
[3] -> https://lore.kernel.org/r/20210825105132.10420-1-kishon@ti.com
Kishon Vijay Abraham I (3):
usb: core: hcd: Modularize HCD stop configuration in usb_stop_hcd()
usb: core: hcd: Add support for deferring roothub registration
xhci: Set HCD flag to defer primary roothub registration
drivers/usb/core/hcd.c | 66 +++++++++++++++++++++++++++--------------
drivers/usb/host/xhci.c | 1 +
include/linux/usb/hcd.h | 2 ++
3 files changed, 47 insertions(+), 22 deletions(-)
--
2.17.1
From: Kishon Vijay Abraham I <hidden> Date: 2021-08-26 11:14:42
No functional change. Since configuration to stop HCD is invoked from
multiple places, group all of them in usb_stop_hcd().
Signed-off-by: Kishon Vijay Abraham I <redacted>
Acked-by: Alan Stern <stern@rowland.harvard.edu>
Tested-by: Chris Chiu <redacted>
---
drivers/usb/core/hcd.c | 39 ++++++++++++++++++++++-----------------
1 file changed, 22 insertions(+), 17 deletions(-)
@@ -2760,6 +2760,26 @@ static void usb_put_invalidate_rhdev(struct usb_hcd *hcd)usb_put_dev(rhdev);}+/**+*usb_stop_hcd-HalttheHCD+*@hcd:theusb_hcdthathastobehalted+*+*Stoptheroot-hubpollingtimerandinvoketheHCD's->stopcallback.+*/+staticvoidusb_stop_hcd(structusb_hcd*hcd)+{+hcd->rh_pollable=0;+clear_bit(HCD_FLAG_POLL_RH,&hcd->flags);+del_timer_sync(&hcd->rh_timer);++hcd->driver->stop(hcd);+hcd->state=HC_STATE_HALT;++/* In case the HCD restarted the timer, stop it again. */+clear_bit(HCD_FLAG_POLL_RH,&hcd->flags);+del_timer_sync(&hcd->rh_timer);+}+/***usb_add_hcd-finishgenericHCDstructureinitializationandregister*@hcd:theusb_hcdstructuretoinitialize
@@ -2946,13 +2966,7 @@ int usb_add_hcd(struct usb_hcd *hcd,returnretval;err_register_root_hub:-hcd->rh_pollable=0;-clear_bit(HCD_FLAG_POLL_RH,&hcd->flags);-del_timer_sync(&hcd->rh_timer);-hcd->driver->stop(hcd);-hcd->state=HC_STATE_HALT;-clear_bit(HCD_FLAG_POLL_RH,&hcd->flags);-del_timer_sync(&hcd->rh_timer);+usb_stop_hcd(hcd);err_hcd_driver_start:if(usb_hcd_is_primary_hcd(hcd)&&hcd->irq>0)free_irq(irqnum,hcd);
@@ -3022,16 +3036,7 @@ void usb_remove_hcd(struct usb_hcd *hcd)*interruptoccurs),butusb_hcd_poll_rh_status()won'tinvoke*thehub_status_data()callback.*/-hcd->rh_pollable=0;-clear_bit(HCD_FLAG_POLL_RH,&hcd->flags);-del_timer_sync(&hcd->rh_timer);--hcd->driver->stop(hcd);-hcd->state=HC_STATE_HALT;--/* In case the HCD restarted the timer, stop it again. */-clear_bit(HCD_FLAG_POLL_RH,&hcd->flags);-del_timer_sync(&hcd->rh_timer);+usb_stop_hcd(hcd);if(usb_hcd_is_primary_hcd(hcd)){if(hcd->irq>0)
From: Kishon Vijay Abraham I <hidden> Date: 2021-08-26 11:14:47
It has been observed with certain PCIe USB cards (like Inateck connected
to AM64 EVM or J7200 EVM) that as soon as the primary roothub is
registered, port status change is handled even before xHC is running
leading to cold plug USB devices not detected. For such cases, registering
both the root hubs along with the second HCD is required. Add support for
deferring roothub registration in usb_add_hcd(), so that both primary and
secondary roothubs are registered along with the second HCD.
Signed-off-by: Kishon Vijay Abraham I <redacted>
Suggested-by: Mathias Nyman <redacted>
Tested-by: Chris Chiu <redacted>
---
drivers/usb/core/hcd.c | 27 ++++++++++++++++++++++-----
include/linux/usb/hcd.h | 2 ++
2 files changed, 24 insertions(+), 5 deletions(-)
@@ -2795,6 +2795,7 @@ int usb_add_hcd(struct usb_hcd *hcd,{intretval;structusb_device*rhdev;+structusb_hcd*shared_hcd;if(!hcd->skip_phy_initialization&&usb_hcd_is_primary_hcd(hcd)){hcd->phy_roothub=usb_phy_roothub_alloc(hcd->self.sysdev);
@@ -2955,18 +2956,34 @@ int usb_add_hcd(struct usb_hcd *hcd,gotoerr_hcd_driver_start;}+/* starting here, usbcore will pay attention to the shared HCD roothub */+shared_hcd=hcd->shared_hcd;+if(!usb_hcd_is_primary_hcd(hcd)&&shared_hcd&&HCD_DEFER_RH_REGISTER(shared_hcd)){+retval=register_root_hub(shared_hcd);+if(retval!=0)+gotoerr_register_shared_root_hub;++if(shared_hcd->uses_new_polling&&HCD_POLL_RH(shared_hcd))+usb_hcd_poll_rh_status(shared_hcd);+}+/* starting here, usbcore will pay attention to this root hub */-retval=register_root_hub(hcd);-if(retval!=0)-gotoerr_register_root_hub;+if(!HCD_DEFER_RH_REGISTER(hcd)){+retval=register_root_hub(hcd);+if(retval!=0)+gotoerr_register_root_hub;-if(hcd->uses_new_polling&&HCD_POLL_RH(hcd))-usb_hcd_poll_rh_status(hcd);+if(hcd->uses_new_polling&&HCD_POLL_RH(hcd))+usb_hcd_poll_rh_status(hcd);+}returnretval;err_register_root_hub:usb_stop_hcd(hcd);+err_register_shared_root_hub:+if(!usb_hcd_is_primary_hcd(hcd)&&shared_hcd&&HCD_DEFER_RH_REGISTER(shared_hcd))+usb_stop_hcd(shared_hcd);err_hcd_driver_start:if(usb_hcd_is_primary_hcd(hcd)&&hcd->irq>0)free_irq(irqnum,hcd);
It has been observed with certain PCIe USB cards (like Inateck connected
to AM64 EVM or J7200 EVM) that as soon as the primary roothub is
registered, port status change is handled even before xHC is running
leading to cold plug USB devices not detected. For such cases, registering
both the root hubs along with the second HCD is required. Add support for
deferring roothub registration in usb_add_hcd(), so that both primary and
secondary roothubs are registered along with the second HCD.
Signed-off-by: Kishon Vijay Abraham I <redacted>
Suggested-by: Mathias Nyman <redacted>
Tested-by: Chris Chiu <redacted>
From: Alan Stern <stern@rowland.harvard.edu> Date: 2021-08-26 15:16:33
On Thu, Aug 26, 2021 at 04:44:25PM +0530, Kishon Vijay Abraham I wrote:
It has been observed with certain PCIe USB cards (like Inateck connected
to AM64 EVM or J7200 EVM) that as soon as the primary roothub is
registered, port status change is handled even before xHC is running
leading to cold plug USB devices not detected. For such cases, registering
both the root hubs along with the second HCD is required. Add support for
deferring roothub registration in usb_add_hcd(), so that both primary and
secondary roothubs are registered along with the second HCD.
Signed-off-by: Kishon Vijay Abraham I <redacted>
Suggested-by: Mathias Nyman <redacted>
Tested-by: Chris Chiu <redacted>
---
@@ -2795,6 +2795,7 @@ int usb_add_hcd(struct usb_hcd *hcd,{intretval;structusb_device*rhdev;+structusb_hcd*shared_hcd;if(!hcd->skip_phy_initialization&&usb_hcd_is_primary_hcd(hcd)){hcd->phy_roothub=usb_phy_roothub_alloc(hcd->self.sysdev);
@@ -2955,18 +2956,34 @@ int usb_add_hcd(struct usb_hcd *hcd,gotoerr_hcd_driver_start;}+/* starting here, usbcore will pay attention to the shared HCD roothub */+shared_hcd=hcd->shared_hcd;+if(!usb_hcd_is_primary_hcd(hcd)&&shared_hcd&&HCD_DEFER_RH_REGISTER(shared_hcd)){+retval=register_root_hub(shared_hcd);+if(retval!=0)+gotoerr_register_shared_root_hub;++if(shared_hcd->uses_new_polling&&HCD_POLL_RH(shared_hcd))+usb_hcd_poll_rh_status(shared_hcd);+}+/* starting here, usbcore will pay attention to this root hub */-retval=register_root_hub(hcd);-if(retval!=0)-gotoerr_register_root_hub;+if(!HCD_DEFER_RH_REGISTER(hcd)){+retval=register_root_hub(hcd);+if(retval!=0)+gotoerr_register_root_hub;-if(hcd->uses_new_polling&&HCD_POLL_RH(hcd))-usb_hcd_poll_rh_status(hcd);+if(hcd->uses_new_polling&&HCD_POLL_RH(hcd))+usb_hcd_poll_rh_status(hcd);+}returnretval;err_register_root_hub:usb_stop_hcd(hcd);+err_register_shared_root_hub:+if(!usb_hcd_is_primary_hcd(hcd)&&shared_hcd&&HCD_DEFER_RH_REGISTER(shared_hcd))+usb_stop_hcd(shared_hcd);
This was added in the wrong place. It needs to come before the
usb_stop_hcd(hcd) call, because usb_start_hcd(hcd) was run before we failed to
register the shared hcd's root hub.
Furthermore, I don't think we need to stop the shared hcd at all. At this
point, the usb_add_hcd() call for the shared hcd has already returned
successfully. The usb_stop_hcd(shared_hcd) callback will be made automatically
when the shared hcd is unregistered later on, after this function returns an
error.
Come to think of it, don't we also need to modify the usb_disconnect(&rhdev)
call in usb_remove_hcd() to handle the case where the root hub wasn't
registered in the first place because it was deferred? At the very least, we
should test the hcd's rh_registered flag before trying to unregister the root
hub.
Alan Stern
quoted hunk
err_hcd_driver_start:
if (usb_hcd_is_primary_hcd(hcd) && hcd->irq > 0)
free_irq(irqnum, hcd);
From: Kishon Vijay Abraham I <hidden> Date: 2021-08-26 11:14:52
Set "HCD_FLAG_DEFER_RH_REGISTER" to hcd->flags in xhci_run() to defer
registering primary roothub in usb_add_hcd(). This will make sure both
primary roothub and secondary roothub will be registered along with the
second HCD. This is required for cold plugged USB devices to be detected
in certain PCIe USB cards (like Inateck USB card connected to AM64 EVM
or J7200 EVM).
Signed-off-by: Kishon Vijay Abraham I <redacted>
Suggested-by: Mathias Nyman <redacted>
Tested-by: Chris Chiu <redacted>
---
drivers/usb/host/xhci.c | 1 +
1 file changed, 1 insertion(+)
Set "HCD_FLAG_DEFER_RH_REGISTER" to hcd->flags in xhci_run() to defer
registering primary roothub in usb_add_hcd(). This will make sure both
primary roothub and secondary roothub will be registered along with the
second HCD. This is required for cold plugged USB devices to be detected
in certain PCIe USB cards (like Inateck USB card connected to AM64 EVM
or J7200 EVM).
Signed-off-by: Kishon Vijay Abraham I <redacted>
Suggested-by: Mathias Nyman <redacted>
Tested-by: Chris Chiu <redacted>
---