On Thu, Jul 26, 2012 at 11:17:38AM -0400, Alan Stern wrote:
> On Thu, 26 Jul 2012, Felipe Balbi wrote:
>
> > Hi,
> >
> > On Wed, Jul 25, 2012 at 02:53:00PM +0200, Rajaram REGUPATHY wrote:
> > > The short_not_ok field is used by class drivers to indicate udc whether short packet is expected during a particular transfer.
> > > In case of mass storage, during command and status phase this field is set as false and set to true during data phase.
> > > musb driver uses this field to decide whether to program DMA for mode1.
> > > This code is essential for musb driver to program DMA.
> > >
> > > Signed-off-by: Balakumar Rajendran <balakumar.rajendran@xxxxxxxxxxxxxx>
> > > Signed-off-by: Rajaram R <ragupathy.rajaram@xxxxxxxxxxxxxx>
> > > ---
> > > drivers/usb/gadget/f_mass_storage.c | 9 +++++++++
> > > 1 files changed, 9 insertions(+), 0 deletions(-)
> > >
> > > diff --git a/drivers/usb/gadget/f_mass_storage.c b/drivers/usb/gadget/f_mass_storage.c
> > > index f67b453..db8bf4a 100644
> > > --- a/drivers/usb/gadget/f_mass_storage.c
> > > +++ b/drivers/usb/gadget/f_mass_storage.c
> > > @@ -878,6 +878,7 @@ static int do_write(struct fsg_common *common)
> > > unsigned int amount;
> > > ssize_t nwritten;
> > > int rc;
> > > + struct usb_gadget *gadget = common->cdev->gadget;
> > >
> > > if (curlun->ro) {
> > > curlun->sense_data = SS_WRITE_PROTECTED; @@ -960,6 +961,8 @@ static int do_write(struct fsg_common *common)
> > > * the bulk-out maxpacket size.
> > > */
> > > set_bulk_out_req_length(common, bh, amount);
> > > + if (!gadget_is_superspeed(gadget))
> > > + bh->outreq->short_not_ok = 1;
> >
> > I'm not sure if this entirely correct. We are now setting req->length
> > aligned to wMaxPacketSize, which means that if we're to receive CBW of
> > 31 bytes, req->length will be set to 512. In this case a short packet
> > _is_ ok. It's the only way for us to know that the transfer has ended.
>
> It's okay. This particular section of code applies only to the data
> phase of a transfer. The CBW is sent in the command phase, and there
> the patch sets short_not_ok to 0.
ok, I didn't read the patch fully, my bad.
--
balbi
Attachment:
signature.asc
Description: Digital signature
