Hi, On Wed, Jul 25, 2012 at 09:10:49PM +0000, Paul Zimmerman wrote: > > From: Alan Stern [mailto:stern@xxxxxxxxxxxxxxxxxxx] > > Sent: Wednesday, July 25, 2012 2:03 PM > > > > On Wed, 25 Jul 2012, Paul Zimmerman wrote: > > > > > > From: Alan Stern [mailto:stern@xxxxxxxxxxxxxxxxxxx] > > > > Sent: Wednesday, July 25, 2012 12:48 PM > > > > To: Paul Zimmerman > > > > Cc: Rajaram R; Michal Nazarewicz; linux-usb@xxxxxxxxxxxxxxx > > > > Subject: RE: usb : mass storage : short_not_ok for non usb3 udc > > > > > > > > On Wed, 25 Jul 2012, Paul Zimmerman wrote: > > > > > > > > > > > > I don't understand this either. What's wrong with setting > > > > > > > > short_not_ok while at SuperSpeed? It shouldn't force the use of a > > > > > > bounce buffer. > > > > > > > > > > > > > > > > > > > > > > I have just brought back some code removed by patch "usb: fix mass > > > > > > > storage gadgets to work with Synopsys UDC". > > > > > > > > > > > > Maybe while bringing it back you can remove the checks for SuperSpeed. > > > > > > Is there any reason to keep them? > > > > > > > > > > Hi Alan, > > > > > > > > > > The problem only arises because at super-speed the MaxPacket size is > > > > > 1024, which is greater than the sector size of 512. So the whole point of > > > > > the patch is to fix the SuperSpeed operation, while allowing high-speed > > > > > devices to operate as they always did. > > > > > > > > Don't the mass-storage drivers already operate correctly at SuperSpeed? > > > > If not, what goes wrong? > > > > > > > > Besides, how could the patch possibly affect SuperSpeed operation? > > > > Everything it does is protected by > > > > > > > > if (!gadget_is_superspeed(gadget)) > > > > > > > > So I'm afraid I don't understand your point. > > > > > > Sorry for not being clear. The original patch: > > > > > > http://www.spinics.net/lists/linux-usb/msg52549.html > > > > > > was required to make mass-storage work with DWC3 at SuperSpeed. > > > However, it apparently broke platforms like musb which depend on > > > short-not-ok being set. The current patch fixes that, by setting > > > short-not-ok if not running SuperSpeed. DWC3 will still work fine > > > with this latest patch, since its problem was only with SuperSpeed. > > > > Okay, that's fine. But it doesn't answer my earlier question: Is there > > really any need for the SuperSpeed checks in this patch? > > > > In other words, will DWC3 continue to work correctly if short_not_ok is > > set? > > Ah, sorry, I missed that. > > I think so, I don't think the DWC3 driver looks at the short-not-ok flag. > But I haven't looked at that driver for a while, so we had better ask > Felipe. > > Felipe? What do you say? We don't use the short_not_ok at all, that's correct. As long as OUT transfers are always aligned on wMaxPacketSize (no matter which speed) DWC3 will continue to work. -- balbi
Attachment:
signature.asc
Description: Digital signature
