]> git.kernelconcepts.de Git - karo-tx-linux.git/commitdiff
usb: gadget: atmel_usba_udc: Expose correct device speed
authorDouglas Gilbert <dgilbert@interlog.com>
Mon, 16 Nov 2015 18:22:08 +0000 (19:22 +0100)
committerFelipe Balbi <balbi@ti.com>
Tue, 17 Nov 2015 17:35:49 +0000 (11:35 -0600)
Following changes that appeared in lk 4.0.0, the gadget udc driver for
some ARM based Atmel SoCs (e.g. at91sam9x5 and sama5d3 families)
incorrectly deduced full-speed USB link speed even when the hardware
had negotiated a high-speed link. The fix is to make sure that the
UDPHS Interrupt Enable Register value does not mask the SPEED bit
in the Interrupt Status Register.

For a mass storage gadget this problem lead to failures when the host
had a USB 3 port with the xhci_hcd driver. If the host was a USB 2
port using the ehci_hcd driver then the mass storage gadget worked
(but probably at a lower speed than it should have).

Signed-off-by: Douglas Gilbert <dgilbert@interlog.com>
Reviewed-by: Boris Brezillon <boris.brezillon@free-electrons.com>
Cc: <stable@vger.kernel.org> #4.0+
Fixes: 9870d895ad87 ("usb: atmel_usba_udc: Mask status with enabled irqs")
Signed-off-by: Felipe Balbi <balbi@ti.com>
drivers/usb/gadget/udc/atmel_usba_udc.c

index f0f2b066ac08366d353b046d3a3a193600977d1d..f92f5aff0dd5e8ece600cc44b5ba6168e3f1bb81 100644 (file)
@@ -1633,7 +1633,7 @@ static irqreturn_t usba_udc_irq(int irq, void *devid)
        spin_lock(&udc->lock);
 
        int_enb = usba_int_enb_get(udc);
-       status = usba_readl(udc, INT_STA) & int_enb;
+       status = usba_readl(udc, INT_STA) & (int_enb | USBA_HIGH_SPEED);
        DBG(DBG_INT, "irq, status=%#08x\n", status);
 
        if (status & USBA_DET_SUSPEND) {