On Tue, 11 Feb 2020, Alan Stern wrote:
> On Tue, 11 Feb 2020, Dan Carpenter wrote:
>
> > Hello USB devs,
> >
> > The patch 7d50195f6c50: "usb: host: Faraday fotg210-hcd driver" from
> > Jul 29, 2013, leads to the following static checker warning:
> >
> > drivers/usb/host/fotg210-hcd.c:3945 iso_stream_init()
> > warn: mask and shift to zero
> >
> > drivers/usb/host/fotg210-hcd.c
> > 3922 static void iso_stream_init(struct fotg210_hcd *fotg210,
> > 3923 struct fotg210_iso_stream *stream, struct usb_device *dev,
> > 3924 int pipe, unsigned interval)
> > 3925 {
> > 3926 u32 buf1;
> > 3927 unsigned epnum, maxp;
> > 3928 int is_input;
> > 3929 long bandwidth;
> > 3930 unsigned multi;
> > 3931
> > 3932 /*
> > 3933 * this might be a "high bandwidth" highspeed endpoint,
> > 3934 * as encoded in the ep descriptor's wMaxPacket field
> > 3935 */
> > 3936 epnum = usb_pipeendpoint(pipe);
> > 3937 is_input = usb_pipein(pipe) ? USB_DIR_IN : 0;
> > 3938 maxp = usb_maxpacket(dev, pipe, !is_input);
> > 3939 if (is_input)
> > 3940 buf1 = (1 << 11);
> > 3941 else
> > 3942 buf1 = 0;
> > 3943
> > 3944 maxp = max_packet(maxp);
> > 3945 multi = hb_mult(maxp);
> > 3946 buf1 |= maxp;
> > 3947 maxp *= multi;
>
> This is pretty clearly a case of mistaken reuse of a local variable.
> The argument to hb_mult() should be the output from usb_maxpacket(),
> not the result of the max_packet() calculation.
Oops, my mistake, sorry. I should have read the source more carefully
before answering.
The max_packet() macro in line 3944 does nothing, since it is only a
mask with 0x07ff and usb_maxpacket() (which calls usb_endpoint_maxp
internally) already performs this mask.
> > 3948
> > 3949 stream->buf0 = cpu_to_hc32(fotg210, (epnum << 8) | dev->devnum);
> > 3950 stream->buf1 = cpu_to_hc32(fotg210, buf1);
> > 3951 stream->buf2 = cpu_to_hc32(fotg210, multi);
> >
> > The problem is these two defines:
> >
> > #define max_packet(wMaxPacketSize) ((wMaxPacketSize) & 0x07ff)
> > #define hb_mult(wMaxPacketSize) (1 + (((wMaxPacketSize) >> 11) & 0x03))
> >
> > 0x07ff >> 11 is always zero so multi is always 1.
> >
> > Should we pass the original value that usb_maxpacket() returned instead
> > of the masked value?
>
> Yes. I suggest introducing a new local variable and using it in place
> of maxp in lines 3938, 3944 (the second occurrence), and 3945.
So the actual right thing to do is to convert the pipe value to a
usb_endpoint_descriptor *, say by way of usb_pipe_endpoint(). Then
maxp and multi can be calculated from the descriptor via
usb_endpoint_maxp() and usb_endpoint_maxp_mult().
Alan Stern
