On Sun, 26 Jul 2015 15:32:54 +0200 Oliver Neukum <oneukum@xxxxxxxx> wrote: > On Fri, 2015-07-24 at 08:48 +0200, Petr Tesarik wrote: > > @@ -872,6 +886,12 @@ static int cp210x_startup(struct usb_serial > > *serial) > > > > usb_set_serial_data(serial, spriv); > > > > + /* Get the 1-byte part number of the cp210x device */ > > + cp210x_control_msg(serial->port[0], CP210X_VENDOR_SPECIFIC, > > + REQTYPE_DEVICE_TO_HOST, CP210X_GET_PARTNUM, > > + &partnum, 1, USB_CTRL_GET_TIMEOUT); > > + spriv->bPartNumber = partnum & 0xFF; > > DMA on the stack. That breaks the cache coherency rules. > You must kmalloc the buffer. I don't understand. While you're right that I copied this part from Sillicon Labs' driver without much thinking, and &spriv->bPartNumber can be used directly, I can't see any DMA on stack. FWIW cp210x_control_msg always allocates a buffer using kcalloc: buf = kcalloc(length, sizeof(__le32), GFP_KERNEL); /* ... */ result = usb_control_msg(serial->dev, usb_rcvctrlpipe(serial->dev, 0), request, requesttype, value, spriv->bInterfaceNumber, buf, size, timeout); Is that what you mean? TIA, Petr T -- To unsubscribe from this list: send the line "unsubscribe linux-usb" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html