Hi, On Tue, Oct 04, 2011 at 12:31:56PM -0700, Paul Zimmerman wrote: > > On Tue, Oct 04, 2011 at 10:44:36AM -0700, Paul Zimmerman wrote: > > > > From: Alan Stern [mailto:stern@xxxxxxxxxxxxxxxxxxx] > > > > Sent: Tuesday, October 04, 2011 6:57 AM > > > > > > > > On Mon, 3 Oct 2011, Paul Zimmerman wrote: > > > > > > > > > Since bulk max packet size must always be a power of 2, we can > > > > > use the 'and' operator instead of modulus. > > > > > > > > Just a quick observation: AFAIK, in wireless USB the bulk maxpacket > > > > size does not have to be a power of 2. On the other hand, if it isn't > > > > then it's not clear that the mass-storage gadgets would work. > > > > > > Drat, I forgot about WUSB and it's dreaded non-power-of-2 max packet > > > sizes. > > > > > > I guess it's better not to apply the patch then, it's probably only > > > a tiny optimization on any halfway modern CPU. > > > > that would mean that for dwc3, we would need to use a chained transfer > > for the remaining bytes, right ? Meaning that we would need a 1024bytes > > throw-away buffer for that, correct ? > > > > I'll start playing with that tomorrow and see how that could be done... > > I guess I could get that finished by v3.3 > > Hi Felipe, > > Not sure what you are referring to here? dwc3 is not wireless USB, so > it will be fine with or without this patch. The patch was only a small > optimization, nothing more. I mean what if we try to read 1536 bytes and we're running at SuperSpeed? Wasn't the purpose of original patch to align req->length to wMaxPacketSize ? -- balbi
Attachment:
signature.asc
Description: Digital signature
