[PATCH v9 0/2] usb: xhci: Add Clear_TT_Buffer

[PATCH v9 0/2] usb: xhci: Add Clear_TT_Buffer

[Jim Lin]

USB 2.0 specification chapter 11.17.5 says "as part of endpoint halt
processing for full-/low-speed endpoints connected via a TT, the host
software must use the Clear_TT_Buffer request to the TT to ensure
that the buffer is not in the busy state".

In our case, a full-speed speaker (ConferenceCam) is behind a high-
speed hub (ConferenceCam Connect), sometimes once we get STALL on a
request we may continue to get STALL with the folllowing requests,
like Set_Interface.

Solution is to invoke usb_hub_clear_tt_buffer() to send
Clear_TT_Buffer request to the hub of the device for the following
Set_Interface requests to the device to get ACK successfully.

The Clear_TT_Buffer request sent to the hub includes the address of
the LS/FS child device in wValue field. usb_hub_clear_tt_buffer()
uses udev->devnum to set the address wValue. This won't work for
devices connected to xHC.

For other host controllers udev->devnum is the same as the address of
the usb device, chosen and set by usb core. With xHC the controller
hardware assigns the address, and won't be the same as devnum.

Here we have two patches.
One is to add devaddr in struct usb_device for
usb_hub_clear_tt_buffer() to use.
Another is to update devaddr in XHCI address_device callback
function and invoke usb_hub_clear_tt_buffer() for halt processing.
 
Signed-off-by: Jim Lin <jilin@nvidia.com>

Jim Lin (2):
  usb: hub : Add devaddr in struct usb_device
  usb: xhci: Add Clear_TT_Buffer

 drivers/usb/core/hub.c       |  4 +++-
 drivers/usb/host/xhci-ring.c | 12 ++++++++++++
 drivers/usb/host/xhci.c      |  1 +
 include/linux/usb.h          |  2 ++
 4 files changed, 18 insertions(+), 1 deletion(-)

-- 
2.1.4

[PATCH v9 1/2] usb: hub : Add devaddr in struct usb_device

[Jim Lin]

The Clear_TT_Buffer request sent to the hub includes the address of
the LS/FS child device in wValue field. usb_hub_clear_tt_buffer()
uses udev->devnum to set the address wValue. This won't work for
devices connected to xHC.

For other host controllers udev->devnum is the same as the address of
the usb device, chosen and set by usb core. With xHC the controller
hardware assigns the address, and won't be the same as devnum.

Here we add devaddr in "struct usb_device" for
usb_hub_clear_tt_buffer() to use.

Signed-off-by: Jim Lin <jilin@nvidia.com>
---
v2: xhci_clear_tt_buffer_complete: add static, shorter indentation
    , remove its claiming in xhci.h
v3: Add description for clearing_tt (xhci.h)
v4: Remove clearing_tt flag because hub_tt_work has hub->tt.lock
    to protect for Clear_TT_Buffer to be run serially.
    Remove xhci_clear_tt_buffer_complete as it's not necessary.
    Same reason as the above.
    Extend usb_hub_clear_tt_buffer parameter
v5: Not extending usb_hub_clear_tt_buffer parameter
    Add description.
v6: Remove unused parameter slot_id from xhci_clear_hub_tt_buffer
v7: Add devaddr field in "struct usb_device"
v8: split as two patches, change type from int to u8 for devaddr.
v9: Use pahole to find place to put devaddr in struct usb_device.
    Remove space between type cast and variable.
    hub.c changed from v8
    clear->devinfo |= (u16) (udev->devaddr << 4);
    to 
    clear->devinfo |= ((u16)udev->devaddr) << 4;
    to solve a problem if devaddr is larger than 16.


 drivers/usb/core/hub.c | 4 +++-
 include/linux/usb.h    | 2 ++
 2 files changed, 5 insertions(+), 1 deletion(-)

diff --git a/drivers/usb/core/hub.c b/drivers/usb/core/hub.c
index 15a2934dc29d..0d4b289be103 100644
--- a/drivers/usb/core/hub.c
+++ b/drivers/usb/core/hub.c
@@ -873,7 +873,7 @@ int usb_hub_clear_tt_buffer(struct urb *urb)
 	/* info that CLEAR_TT_BUFFER needs */
 	clear->tt = tt->multi ? udev->ttport : 1;
 	clear->devinfo = usb_pipeendpoint (pipe);
-	clear->devinfo |= udev->devnum << 4;
+	clear->devinfo |= ((u16)udev->devaddr) << 4;
 	clear->devinfo |= usb_pipecontrol(pipe)
 			? (USB_ENDPOINT_XFER_CONTROL << 11)
 			: (USB_ENDPOINT_XFER_BULK << 11);
@@ -2125,6 +2125,8 @@ static void update_devnum(struct usb_device *udev, int devnum)
 	/* The address for a WUSB device is managed by wusbcore. */
 	if (!udev->wusb)
 		udev->devnum = devnum;
+	if (!udev->devaddr)
+		udev->devaddr = (u8)devnum;
 }
 
 static void hub_free_dev(struct usb_device *udev)
diff --git a/include/linux/usb.h b/include/linux/usb.h
index 4229eb74bd2c..5641965e44a1 100644
--- a/include/linux/usb.h
+++ b/include/linux/usb.h
@@ -566,6 +566,7 @@ struct usb3_lpm_parameters {
  * @tt: Transaction Translator info; used with low/full speed dev, highspeed hub
  * @ttport: device port on that tt hub
  * @toggle: one bit for each endpoint, with ([0] = IN, [1] = OUT) endpoints
+ * @devaddr: device address, XHCI: assigned by HW, others: same as devnum
  * @parent: our hub, unless we're the root
  * @bus: bus we're part of
  * @ep0: endpoint 0 data (default control pipe)
@@ -643,6 +644,7 @@ struct usb_device {
 	int		ttport;
 
 	unsigned int toggle[2];
+	u8 devaddr;
 
 	struct usb_device *parent;
 	struct usb_bus *bus;
-- 
2.1.4

[PATCH v9 2/2] usb: xhci: Add Clear_TT_Buffer

[Jim Lin]

USB 2.0 specification chapter 11.17.5 says "as part of endpoint halt
processing for full-/low-speed endpoints connected via a TT, the host
software must use the Clear_TT_Buffer request to the TT to ensure
that the buffer is not in the busy state".

In our case, a full-speed speaker (ConferenceCam) is behind a high-
speed hub (ConferenceCam Connect), sometimes once we get STALL on a
request we may continue to get STALL with the folllowing requests,
like Set_Interface.

Here we update udev->devaddr in address_device callback function
(this USB device address is assigned by XHCI controller) and invoke
usb_hub_clear_tt_buffer() to send Clear_TT_Buffer request to the hub
of the device for the following Set_Interface requests to the device
to get ACK successfully.

Signed-off-by: Jim Lin <jilin@nvidia.com>
---
v2: xhci_clear_tt_buffer_complete: add static, shorter indentation
    , remove its claiming in xhci.h
v3: Add description for clearing_tt (xhci.h)
v4: Remove clearing_tt flag because hub_tt_work has hub->tt.lock
    to protect for Clear_TT_Buffer to be run serially.
    Remove xhci_clear_tt_buffer_complete as it's not necessary.
    Same reason as the above.
    Extend usb_hub_clear_tt_buffer parameter
v5: Not extending usb_hub_clear_tt_buffer parameter
    Add description.
v6: Remove unused parameter slot_id from xhci_clear_hub_tt_buffer
v7: Add devaddr field in "struct usb_device"
v8: split as two patches
v9: no change

 drivers/usb/host/xhci-ring.c | 12 ++++++++++++
 drivers/usb/host/xhci.c      |  1 +
 2 files changed, 13 insertions(+)

diff --git a/drivers/usb/host/xhci-ring.c b/drivers/usb/host/xhci-ring.c
index 9215a28dad40..739737faf752 100644
--- a/drivers/usb/host/xhci-ring.c
+++ b/drivers/usb/host/xhci-ring.c
@@ -1786,6 +1786,17 @@ struct xhci_segment *trb_in_td(struct xhci_hcd *xhci,
 	return NULL;
 }
 
+static void xhci_clear_hub_tt_buffer(struct xhci_hcd *xhci, struct xhci_td *td)
+{
+	/*
+	 * As part of low/full-speed endpoint-halt processing
+	 * we must clear the TT buffer (USB 2.0 specification 11.17.5).
+	 */
+	if (td->urb->dev->tt && !usb_pipeint(td->urb->pipe) &&
+	    (td->urb->dev->tt->hub != xhci_to_hcd(xhci)->self.root_hub))
+		usb_hub_clear_tt_buffer(td->urb);
+}
+
 static void xhci_cleanup_halted_endpoint(struct xhci_hcd *xhci,
 		unsigned int slot_id, unsigned int ep_index,
 		unsigned int stream_id, struct xhci_td *td,
@@ -1804,6 +1815,7 @@ static void xhci_cleanup_halted_endpoint(struct xhci_hcd *xhci,
 	if (reset_type == EP_HARD_RESET) {
 		ep->ep_state |= EP_HARD_CLEAR_TOGGLE;
 		xhci_cleanup_stalled_ring(xhci, ep_index, stream_id, td);
+		xhci_clear_hub_tt_buffer(xhci, td);
 	}
 	xhci_ring_cmd_db(xhci);
 }
diff --git a/drivers/usb/host/xhci.c b/drivers/usb/host/xhci.c
index 7fa58c99f126..68b393e5a453 100644
--- a/drivers/usb/host/xhci.c
+++ b/drivers/usb/host/xhci.c
@@ -4096,6 +4096,7 @@ static int xhci_setup_device(struct usb_hcd *hcd, struct usb_device *udev,
 	/* Zero the input context control for later use */
 	ctrl_ctx->add_flags = 0;
 	ctrl_ctx->drop_flags = 0;
+	udev->devaddr = (u8)(le32_to_cpu(slot_ctx->dev_state) & DEV_ADDR_MASK);
 
 	xhci_dbg_trace(xhci, trace_xhci_dbg_address,
 		       "Internal device address = %d",
-- 
2.1.4

Re: [PATCH v9 1/2] usb: hub : Add devaddr in struct usb_device

[Alan Stern]

On Fri, 10 May 2019, Jim Lin wrote:

> The Clear_TT_Buffer request sent to the hub includes the address of
> the LS/FS child device in wValue field. usb_hub_clear_tt_buffer()
> uses udev->devnum to set the address wValue. This won't work for
> devices connected to xHC.
> 
> For other host controllers udev->devnum is the same as the address of
> the usb device, chosen and set by usb core. With xHC the controller
> hardware assigns the address, and won't be the same as devnum.
> 
> Here we add devaddr in "struct usb_device" for
> usb_hub_clear_tt_buffer() to use.
> 
> Signed-off-by: Jim Lin <jilin@nvidia.com>
> ---
> v2: xhci_clear_tt_buffer_complete: add static, shorter indentation
>     , remove its claiming in xhci.h
> v3: Add description for clearing_tt (xhci.h)
> v4: Remove clearing_tt flag because hub_tt_work has hub->tt.lock
>     to protect for Clear_TT_Buffer to be run serially.
>     Remove xhci_clear_tt_buffer_complete as it's not necessary.
>     Same reason as the above.
>     Extend usb_hub_clear_tt_buffer parameter
> v5: Not extending usb_hub_clear_tt_buffer parameter
>     Add description.
> v6: Remove unused parameter slot_id from xhci_clear_hub_tt_buffer
> v7: Add devaddr field in "struct usb_device"
> v8: split as two patches, change type from int to u8 for devaddr.
> v9: Use pahole to find place to put devaddr in struct usb_device.
>     Remove space between type cast and variable.
>     hub.c changed from v8
>     clear->devinfo |= (u16) (udev->devaddr << 4);
>     to 
>     clear->devinfo |= ((u16)udev->devaddr) << 4;
>     to solve a problem if devaddr is larger than 16.
> 
> 
>  drivers/usb/core/hub.c | 4 +++-
>  include/linux/usb.h    | 2 ++
>  2 files changed, 5 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/usb/core/hub.c b/drivers/usb/core/hub.c
> index 15a2934dc29d..0d4b289be103 100644
> --- a/drivers/usb/core/hub.c
> +++ b/drivers/usb/core/hub.c
> @@ -873,7 +873,7 @@ int usb_hub_clear_tt_buffer(struct urb *urb)
>  	/* info that CLEAR_TT_BUFFER needs */
>  	clear->tt = tt->multi ? udev->ttport : 1;
>  	clear->devinfo = usb_pipeendpoint (pipe);
> -	clear->devinfo |= udev->devnum << 4;
> +	clear->devinfo |= ((u16)udev->devaddr) << 4;
>  	clear->devinfo |= usb_pipecontrol(pipe)
>  			? (USB_ENDPOINT_XFER_CONTROL << 11)
>  			: (USB_ENDPOINT_XFER_BULK << 11);
> @@ -2125,6 +2125,8 @@ static void update_devnum(struct usb_device *udev, int devnum)
>  	/* The address for a WUSB device is managed by wusbcore. */
>  	if (!udev->wusb)
>  		udev->devnum = devnum;
> +	if (!udev->devaddr)
> +		udev->devaddr = (u8)devnum;

Whether or not to put a space after the ')' in a typecast is a matter 
of personal choice; CodingStyle doesn't say anything about it.  Quite 
often the space is present.  IIRC, that is how K&R does it.  But it 
doesn't really matter.

>  }
>  
>  static void hub_free_dev(struct usb_device *udev)
> diff --git a/include/linux/usb.h b/include/linux/usb.h
> index 4229eb74bd2c..5641965e44a1 100644
> --- a/include/linux/usb.h
> +++ b/include/linux/usb.h
> @@ -566,6 +566,7 @@ struct usb3_lpm_parameters {
>   * @tt: Transaction Translator info; used with low/full speed dev, highspeed hub
>   * @ttport: device port on that tt hub
>   * @toggle: one bit for each endpoint, with ([0] = IN, [1] = OUT) endpoints
> + * @devaddr: device address, XHCI: assigned by HW, others: same as devnum
>   * @parent: our hub, unless we're the root
>   * @bus: bus we're part of
>   * @ep0: endpoint 0 data (default control pipe)
> @@ -643,6 +644,7 @@ struct usb_device {
>  	int		ttport;
>  
>  	unsigned int toggle[2];
> +	u8 devaddr;

This is another lousy place to put the new value.  A much better
position would be immediately after "u8 level".

>  	struct usb_device *parent;
>  	struct usb_bus *bus;

You should move into this patch the code added to xhci.c to assign
udev->devaddr.  Without it, this patch is incomplete: devaddr will not
contain the correct value for devices on a USB-3 bus.  And it doesn't
belong in the 2/2 patch -- that patch is all about making xhci-hcd call
usb_clear_tt_buffer().

Alan Stern

Re: [PATCH v9 2/2] usb: xhci: Add Clear_TT_Buffer

[Alan Stern]

On Fri, 10 May 2019, Jim Lin wrote:

> USB 2.0 specification chapter 11.17.5 says "as part of endpoint halt
> processing for full-/low-speed endpoints connected via a TT, the host
> software must use the Clear_TT_Buffer request to the TT to ensure
> that the buffer is not in the busy state".
> 
> In our case, a full-speed speaker (ConferenceCam) is behind a high-
> speed hub (ConferenceCam Connect), sometimes once we get STALL on a
> request we may continue to get STALL with the folllowing requests,
> like Set_Interface.
> 
> Here we update udev->devaddr in address_device callback function
> (this USB device address is assigned by XHCI controller) and invoke
> usb_hub_clear_tt_buffer() to send Clear_TT_Buffer request to the hub
> of the device for the following Set_Interface requests to the device
> to get ACK successfully.
> 
> Signed-off-by: Jim Lin <jilin@nvidia.com>
> ---
> v2: xhci_clear_tt_buffer_complete: add static, shorter indentation
>     , remove its claiming in xhci.h
> v3: Add description for clearing_tt (xhci.h)
> v4: Remove clearing_tt flag because hub_tt_work has hub->tt.lock
>     to protect for Clear_TT_Buffer to be run serially.
>     Remove xhci_clear_tt_buffer_complete as it's not necessary.
>     Same reason as the above.
>     Extend usb_hub_clear_tt_buffer parameter
> v5: Not extending usb_hub_clear_tt_buffer parameter
>     Add description.
> v6: Remove unused parameter slot_id from xhci_clear_hub_tt_buffer
> v7: Add devaddr field in "struct usb_device"
> v8: split as two patches
> v9: no change
> 
>  drivers/usb/host/xhci-ring.c | 12 ++++++++++++
>  drivers/usb/host/xhci.c      |  1 +
>  2 files changed, 13 insertions(+)
> 
> diff --git a/drivers/usb/host/xhci-ring.c b/drivers/usb/host/xhci-ring.c
> index 9215a28dad40..739737faf752 100644
> --- a/drivers/usb/host/xhci-ring.c
> +++ b/drivers/usb/host/xhci-ring.c
> @@ -1786,6 +1786,17 @@ struct xhci_segment *trb_in_td(struct xhci_hcd *xhci,
>  	return NULL;
>  }
>  
> +static void xhci_clear_hub_tt_buffer(struct xhci_hcd *xhci, struct xhci_td *td)
> +{
> +	/*
> +	 * As part of low/full-speed endpoint-halt processing
> +	 * we must clear the TT buffer (USB 2.0 specification 11.17.5).
> +	 */
> +	if (td->urb->dev->tt && !usb_pipeint(td->urb->pipe) &&
> +	    (td->urb->dev->tt->hub != xhci_to_hcd(xhci)->self.root_hub))
> +		usb_hub_clear_tt_buffer(td->urb);
> +}
> +
>  static void xhci_cleanup_halted_endpoint(struct xhci_hcd *xhci,
>  		unsigned int slot_id, unsigned int ep_index,
>  		unsigned int stream_id, struct xhci_td *td,
> @@ -1804,6 +1815,7 @@ static void xhci_cleanup_halted_endpoint(struct xhci_hcd *xhci,
>  	if (reset_type == EP_HARD_RESET) {
>  		ep->ep_state |= EP_HARD_CLEAR_TOGGLE;
>  		xhci_cleanup_stalled_ring(xhci, ep_index, stream_id, td);
> +		xhci_clear_hub_tt_buffer(xhci, td);
>  	}
>  	xhci_ring_cmd_db(xhci);
>  }

How come there's no clear_tt_buffer_complete() callback?  Without it,
xhci-hcd won't know when the TT buffer has been cleared, and so it
won't know when to restart the endpoint ring for the halted endpoint.  

Note: Restarting the endpoint ring before the TT buffer has been 
cleared is not safe.

Alan Stern

> diff --git a/drivers/usb/host/xhci.c b/drivers/usb/host/xhci.c
> index 7fa58c99f126..68b393e5a453 100644
> --- a/drivers/usb/host/xhci.c
> +++ b/drivers/usb/host/xhci.c
> @@ -4096,6 +4096,7 @@ static int xhci_setup_device(struct usb_hcd *hcd, struct usb_device *udev,
>  	/* Zero the input context control for later use */
>  	ctrl_ctx->add_flags = 0;
>  	ctrl_ctx->drop_flags = 0;
> +	udev->devaddr = (u8)(le32_to_cpu(slot_ctx->dev_state) & DEV_ADDR_MASK);
>  
>  	xhci_dbg_trace(xhci, trace_xhci_dbg_address,
>  		       "Internal device address = %d",
>