On 02/02/2010 06:42 AM, Michael Buesch wrote:
> Hi,
>
> I've got a buggy device that needs a special workaround to be usable under
> host-usb access. The device really doesn't like being reset via
> USBDEVFS_RESET. It
> immediatenly locks up the device firmware or whatever. It won't respond
> properly anymore.
> With the following patch it works fine, though.
>
What about the USBDEVFS_RESET in usb_host_open? Does that have an impact?
For some USB keys I have had to add an additional reset prior to
claiming interfaces:
diff --git a/usb-linux.c b/usb-linux.c
index 1aaa595..092e75c 100644
--- a/usb-linux.c
+++ b/usb-linux.c
@@ -906,6 +906,9 @@ static int usb_host_open(USBHostDevice *dev, int
bus_num,
#endif
+ /* some keys require a reset before the getconfig */
+ ioctl(fd, USBDEVFS_RESET);
+
/*
* Initial configuration is -1 which makes us claim first
* available config. We used to start with 1, which does not
David Ahern
> So I was wondering what the accepted way was to get these quirks upstream
> into the qemu
> source tree. Is usb-linux.c the correct place, or should we put the quirk into
> a different place?
>
> ---
> usb-linux.c | 4 ++++
> 1 file changed, 4 insertions(+)
>
> --- qemu.orig/usb-linux.c
> +++ qemu/usb-linux.c
> @@ -389,6 +389,10 @@ static void usb_host_handle_reset(USBDev
>
> dprintf("husb: reset device %u.%u\n", s->bus_num, s->addr);
>
> + if (((s->descr[8] << 8) | s->descr[9]) == 0x2471 &&
> + ((s->descr[10] << 8) | s->descr[11]) == 0x0853)
> + return;
> +
> ioctl(s->fd, USBDEVFS_RESET);
>
> usb_host_claim_interfaces(s, s->configuration);
>
