Provide a hook to check port init status
Subsystems (2)
cover Provide a hook to check port init status
Currently, only usbcore knows port init status even if the result is bad, so it will keep doing USB enumeration and other drivers can't do anything. Add the hook to let other drivers know the status and do error handling.
Ray Chi (2):
usb: core: add a hook to check port init status
usb: xhci: add check_init_status hook support
drivers/usb/core/hub.c | 14 ++++++++++++++ drivers/usb/host/xhci.c | 17 +++++++++++++++++ drivers/usb/host/xhci.h | 1 + include/linux/usb/hcd.h | 8 ++++++++ 4 files changed, 40 insertions(+)
Signature · 1 line
2.37.1.595.g718a3a8f04-goog
PATCH 1/2 usb: core: add a hook to check port init status
This patch add a hook to check the port init status. Currently, only usbcore knows port init status even if the result is bad. It will cause a USB host keep doing USB enumeration for a long time when the USB host connects to a broken USB accessory.
The hc_driver could use the hook to know port init status and do possible error handling according to platform requirements or limitations.
--- drivers/usb/core/hub.c | 14 ++++++++++++++ include/linux/usb/hcd.h | 8 ++++++++ 2 files changed, 22 insertions(+)
drivers/usb/core/hub.c +14 -0
@@ -4659,6 +4659,16 @@ static int hub_enable_device(struct usb_device *udev)return hcd->driver->enable_device (hcd, udev); } +static int hub_port_check_init_status (struct usb_device *udev,int r) +{ +struct usb_hcd *hcd =bus_to_hcd (udev->bus); + +if (!hcd->driver->check_init_status) +return 0 ; + +return hcd->driver->check_init_status (hcd, udev, r); +} +/* Reset device, (re)assign address, get device descriptor. * Device connection must be stable, no more debouncing needed. * Returns device in USB_STATE_ADDRESS, except on error.
@@ -4855,6 +4865,10 @@ hub_port_init(struct usb_hub *hub, struct usb_device *udev, int port1, buf->bMaxPacketSize0;kfree (buf); + retval =hub_port_check_init_status (udev, r); +if (retval <0 ) + goto fail; + retval =hub_port_reset (hub, port1, udev, delay, false);if (retval <0 )/* error or disconnect */ goto fail;
include/linux/usb/hcd.h +8 -0
@@ -297,6 +297,14 @@ struct hc_driver { gfp_t mem_flags);void (*unmap_urb_for_dma)(struct usb_hcd *hcd,struct urb *urb); +/* +* (optional) HCD could get the information of GET_DESCRIPTOR by this hook. +* In general, it is not necessary unless the enumeration takes long +* time to do. The host controller could know the enumeration status by +* this hook and do some error handlings. +*/ +int (*check_init_status)(struct usb_hcd *hcd,struct usb_device *udev,int r); +/* hw synch, freeing endpoint resources that urb_dequeue can't */ void (*endpoint_disable)(struct usb_hcd *hcd,struct usb_host_endpoint *ep);
Signature · 1 line
2.37.1.595.g718a3a8f04-goog
Re: [PATCH 1/2] usb: core: add a hook to check port init status
Quoted 20 lines · Ray Chi
This patch add a hook to check the port init status. Currently, only usbcore knows port init status even if the result is bad. It will cause a USB host keep doing USB enumeration for a long time when the USB host connects to a broken USB accessory.
The
hc_drivercould use the hook to know port init status and do possible error handling according to platform requirements or limitations.Signed-off-by Ray Chi <[email protected]>--- drivers/usb/core/hub.c | 14 ++++++++++++++ include/linux/usb/hcd.h | 8 ++++++++ 2 files changed, 22 insertions(+)drivers/usb/core/hub.c +0
-0
drivers/usb/core/hub.c hub_port_init(struct usb_hub *hub, struct usb_device *udev, int port1, · line 4870
Quoted 6 lines
drivers/usb/core/hub.c +3
-0@@ -4855,6 +4865,10 @@ hub_port_init(struct usb_hub *hub, struct usb_device *udev, int port1, buf->bMaxPacketSize0;kfree (buf); + retval =hub_port_check_init_status (udev, r); +if (retval <0 ) + goto fail;
For future reference, you should be aware that this code won't get executed if do_new_scheme is false.
Alan Stern
PATCH 2/2 usb: xhci: add check_init_status hook support
In reply to Ray ChiIn general, xHCI didn't do anything for port initialization. However, there are some requirement or limitation on various platforms, so vendors need to do some error handlings if the device connected to a broken USB accessory.
This patch also add the hook to xhci_driver_overrides so that vendors can add their specific protection easily if needed.
--- drivers/usb/host/xhci.c | 17 +++++++++++++++++ drivers/usb/host/xhci.h | 1 + 2 files changed, 18 insertions(+)
drivers/usb/host/xhci.c +17 -0
@@ -4358,6 +4358,20 @@ static int xhci_enable_device(struct usb_hcd *hcd, struct usb_device *udev)return xhci_setup_device (hcd, udev,SETUP_CONTEXT_ONLY ); } +/* +* The function could get the status of port initialization. +*/ +static int xhci_check_init_status (struct usb_hcd *hcd,struct usb_device *udev,int r) +{ +/* +* In general, this function is not necessory. Some platforms may +* need doing error handling when the port initialization takes a +* long time to do. The device can use the override callback to +* do specific handlings. +*/ +return 0 ; +} +/* * Transfer the port index into real index in the HW port status * registers. Caculate offset between the port's PORTSC register
@@ -5455,6 +5469,7 @@ static const struct hc_driver xhci_hc_driver = { .enable_device = xhci_enable_device, .update_hub_device = xhci_update_hub_device, .reset_device = xhci_discover_or_reset_device, + .check_init_status = xhci_check_init_status,/* * scheduling support
@@ -5503,6 +5518,8 @@ void xhci_init_driver(struct hc_driver *drv, drv->check_bandwidth = over->check_bandwidth;if (over->reset_bandwidth) drv->reset_bandwidth = over->reset_bandwidth; +if (over->check_init_status) + drv->check_init_status = over->check_init_status; } }EXPORT_SYMBOL_GPL (xhci_init_driver);
drivers/usb/host/xhci.h +1 -0
@@ -1944,6 +1944,7 @@ struct xhci_driver_overrides { struct usb_host_endpoint *ep);int (*check_bandwidth)(struct usb_hcd *,struct usb_device *);void (*reset_bandwidth)(struct usb_hcd *,struct usb_device *); +int (*check_init_status)(struct usb_hcd *hcd,struct usb_device *udev,int r); };#define XHCI_CFC_DELAY 10
Signature · 1 line
2.37.1.595.g718a3a8f04-goog
Greg KH
maintainerRe: [PATCH 2/2] usb: xhci: add check_init_status hook support
Quoted 25 lines · Ray Chi
In general, xHCI didn't do anything for port initialization. However, there are some requirement or limitation on various platforms, so vendors need to do some error handlings if the device connected to a broken USB accessory.
This patch also add the hook to
xhci_driver_overridesso that vendors can add their specific protection easily if needed.Signed-off-by Ray Chi <[email protected]>--- drivers/usb/host/xhci.c | 17 +++++++++++++++++ drivers/usb/host/xhci.h | 1 + 2 files changed, 18 insertions(+)drivers/usb/host/xhci.c +3
-0@@ -4358,6 +4358,20 @@ static int xhci_enable_device(struct usb_hcd *hcd, struct usb_device *udev)return xhci_setup_device (hcd, udev,SETUP_CONTEXT_ONLY ); } +/* +* The function could get the status of port initialization. +*/
drivers/usb/host/xhci.c static int xhci_enable_device(struct usb_hcd *hcd, struct usb_device *udev) · line 4373
Quoted 10 lines
drivers/usb/host/xhci.c +10
-0+static int xhci_check_init_status (struct usb_hcd *hcd,struct usb_device *udev,int r) +{ +/* +* In general, this function is not necessory. Some platforms may +* need doing error handling when the port initialization takes a +* long time to do. The device can use the override callback to +* do specific handlings. +*/ +return 0 ; +}
For obvious technical and legal reasons, we are not allowed to add "hooks" to the kernel where there are no in-kernel users. Nor would you want us to do so.
So I really do not understand this patch series at all.
What driver wants to do odd things here? What needs to happen that the in-tree drivers are not doing properly? Why not get the needed fixes in the in-kernel drivers instead of trying to add random hooks that some out-of-tree code would use instead.
confused,
greg k-h
Re: [PATCH 2/2] usb: xhci: add check_init_status hook support
Quoted 33 lines · Greg KH
Quoted 31 lines · Ray Chi
In general, xHCI didn't do anything for port initialization. However, there are some requirement or limitation on various platforms, so vendors need to do some error handlings if the device connected to a broken USB accessory.
This patch also add the hook to
xhci_driver_overridesso that vendors can add their specific protection easily if needed.Signed-off-by Ray Chi <[email protected]>--- drivers/usb/host/xhci.c | 17 +++++++++++++++++ drivers/usb/host/xhci.h | 1 + 2 files changed, 18 insertions(+)drivers/usb/host/xhci.c +9
-0@@ -4358,6 +4358,20 @@ static int xhci_enable_device(struct usb_hcd *hcd, struct usb_device *udev)return xhci_setup_device (hcd, udev,SETUP_CONTEXT_ONLY ); } +/* +* The function could get the status of port initialization. +*/ +static int xhci_check_init_status (struct usb_hcd *hcd,struct usb_device *udev,int r) +{ +/* +* In general, this function is not necessory. Some platforms may +* need doing error handling when the port initialization takes a +* long time to do. The device can use the override callback to
Quoted 9 lines
drivers/usb/host/xhci.c static int xhci_enable_device(struct usb_hcd *hcd, struct usb_device *udev) · line 4373
Quoted 4 lines
drivers/usb/host/xhci.c +4
-0+* do specific handlings. +*/ +return 0 ; +}For obvious technical and legal reasons, we are not allowed to add "hooks" to the kernel where there are no in-kernel users. Nor would you want us to do so.
Agree on this. I am trying another way to achieve the same goal.
Quoted 10 lines
So I really do not understand this patch series at all.
What driver wants to do odd things here? What needs to happen that the in-tree drivers are not doing properly? Why not get the needed fixes in the in-kernel drivers instead of trying to add random hooks that some out-of-tree code would use instead.
confused,
greg k-h
I will prepare a new commit to do it.
Thanks,
Ray