Hi,
Peter Chen <hzpeterchen@xxxxxxxxx> writes:
> On Thu, Nov 03, 2016 at 11:27:40AM +0200, Felipe Balbi wrote:
>>
>> Hi,
>>
>> Peter Chen <hzpeterchen@xxxxxxxxx> writes:
>> >> Peter Chen <peter.chen@xxxxxxx> writes:
>> >> > For isoc endpoint descriptor, the wMaxPacketSize is not real max packet
>> >> > size (see Table 9-13. Standard Endpoint Descriptor, USB 2.0 specifcation),
>> >> > it may contain the number of packet, so the real max packet should be
>> >> > ep->desc->wMaxPacketSize && 0x7ff.
>> >> >
>> >> > Cc: Felipe F. Tonello <eu@xxxxxxxxxxxxxxxxx>
>> >> > Cc: Felipe Balbi <felipe.balbi@xxxxxxxxxxxxxxx>
>> >> > Fixes: 16b114a6d797 ("usb: gadget: fix usb_ep_align_maybe
>> >> > endianness and new usb_ep_aligna")
>> >> >
>> >> > Signed-off-by: Peter Chen <peter.chen@xxxxxxx>
>> >> > ---
>> >> > include/linux/usb/gadget.h | 5 ++++-
>> >> > 1 file changed, 4 insertions(+), 1 deletion(-)
>> >> >
>> >> > diff --git a/include/linux/usb/gadget.h b/include/linux/usb/gadget.h
>> >> > index 8e81f9e..cce2db6 100644
>> >> > --- a/include/linux/usb/gadget.h
>> >> > +++ b/include/linux/usb/gadget.h
>> >> > @@ -429,7 +429,10 @@ static inline struct usb_gadget *dev_to_usb_gadget(struct device *dev)
>> >> > */
>> >> > static inline size_t usb_ep_align(struct usb_ep *ep, size_t len)
>> >> > {
>> >> > - return round_up(len, (size_t)le16_to_cpu(ep->desc->wMaxPacketSize));
>> >> > + int max_packet_size = (size_t)le16_to_cpu(ep->desc->wMaxPacketSize)
>> >> > + && 0x7ff;
>> ^^
>> hehe, I just noticed this bug. Did you really test this patch? :-)
>>
>
> Any problems you find? Without this patch, the usbtest can't be ran
> for high bandwidth isoc transfer.
wMaxPacketSize && 0x7ff is always 1 unless wMaxPacketSize is zero, which
is invalid anyway.
>> >> > +
>> >> > + return round_up(len, max_packet_size);
>> >> > }
>> >>
>> >> care to update this to use new usb_endpoint_maxp()?
>> >>
>> >
>> > But it is a bug existed at v4.9, the usb_ep_align returns wrong request
>> > length for isoc. I found this bug by running usbtest.
>>
>> keep your & (after fixing it), but use usb_endpoint_maxp(). Later we can
>> just remove the unnecessary & operation.
>>
>> --
>> balbi
>
> I will use usb_endpoint_maxp, but I still not find the problem for this
> patch.
Logical vs Bitwise operator
--
balbi
Attachment:
signature.asc
Description: PGP signature
