On Mon, May 02, 2022 at 09:07:58AM +0200, Jose Ignacio Tornos Martinez wrote:
> Bluetooth Dongles with CSR chip (i.e. USB Bluetooth V4.0 Dongle by
> Trust) hang when they are unbound from 'unbind' sysfs entry and
> can not be bound again.
>
> The reason is CSR chip hangs when usb configuration command with
> index 0 (used to unconfigure) is sent during disconnection.
>
> To avoid this unwanted result, it is necessary not to send this
> command for CSR chip when usb device is unbound, so a new quirk
> has been created for this device.
>
> Athough device is not unconfigured, it is better to avoid device
> hanging to be able to operate. Even bluetooth can be previously
> turned off.
> On the other hand, this is not important if usb device is going to
> be bound again (normal behavior), i.e. with usbip.
> ---
> drivers/usb/core/generic.c | 3 ++-
> drivers/usb/core/quirks.c | 3 +++
> include/linux/usb/quirks.h | 3 +++
> 3 files changed, 8 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/usb/core/generic.c b/drivers/usb/core/generic.c
> index 740342a2812a..ea770b83d876 100644
> --- a/drivers/usb/core/generic.c
> +++ b/drivers/usb/core/generic.c
> @@ -22,6 +22,7 @@
> #include <linux/usb.h>
> #include <linux/usb/hcd.h>
> #include <uapi/linux/usb/audio.h>
> +#include <linux/usb/quirks.h>
> #include "usb.h"
>
> static inline const char *plural(int n)
> @@ -256,7 +257,7 @@ void usb_generic_driver_disconnect(struct usb_device *udev)
>
> /* if this is only an unbind, not a physical disconnect, then
> * unconfigure the device */
> - if (udev->actconfig)
> + if (!(udev->quirks & USB_QUIRK_SKIP_UNCONFIGURE) && udev->actconfig)
The usual programming convention is that when an "if" statement tests
more than one condition, the test which is more likely to fail should
come first. In this case, it is more likely that udev->actconfig is
non-NULL than that the USB_QUIRK_SKIP_UNCONFIGURE flag is set, because
the quirk flag affects only one type of device whereas udev->actconfig
can be non-NULL on any device. So the test of udev->actconfig should
come first.
> usb_set_configuration(udev, -1);
> }
>
> diff --git a/drivers/usb/core/quirks.c b/drivers/usb/core/quirks.c
> index d3c14b5ed4a1..13989629d743 100644
> --- a/drivers/usb/core/quirks.c
> +++ b/drivers/usb/core/quirks.c
> @@ -510,6 +510,9 @@ static const struct usb_device_id usb_quirk_list[] = {
> /* INTEL VALUE SSD */
> { USB_DEVICE(0x8086, 0xf1a5), .driver_info = USB_QUIRK_RESET_RESUME },
>
> + /* CSR Bluetooth */
> + { USB_DEVICE(0x0a12, 0x0001), .driver_info = USB_QUIRK_SKIP_UNCONFIGURE },
> +
> { } /* terminating entry must be last */
> };
Did you want people to be able to test this quirk interactively? If you
do, you should add a flag character for this quirk to quirks_param_set()
and document it. But if you don't think people will need to test the
new quirk then you don't need to add anything else.
> diff --git a/include/linux/usb/quirks.h b/include/linux/usb/quirks.h
> index eeb7c2157c72..84ab0a369931 100644
> --- a/include/linux/usb/quirks.h
> +++ b/include/linux/usb/quirks.h
> @@ -72,4 +72,7 @@
> /* device has endpoints that should be ignored */
> #define USB_QUIRK_ENDPOINT_IGNORE BIT(15)
>
> +/* device doesn't support unconfigure when unbound. */
> +#define USB_QUIRK_SKIP_UNCONFIGURE BIT(16)
The comment isn't right. In fact, the device doesn't ever support
unconfigure. Whether it is bound or not makes no difference.
Alan Stern
