> From: Sarah Sharp [mailto:sarah.a.sharp@xxxxxxxxxxxxxxx] > > On Thu, Oct 31, 2013 at 04:16:14PM -0000, David Laight wrote: > > Code was recently added to the xhci driver to dynamically add extra ring > > segments if there is insufficient space for all the fragments of the > > URB being sent. > > > > ISTM that it would be better to just queue the URB (on the ring) > > and process it when space becomes available. > > Hmm, better why? I'm reluctant to touch code that Just Works (TM). > Especially anything having do with the xHCI ring code, since it's quite > complex. I was thinking that queueing URB would make it possible to significantly simplify the ring code - especially since it could be changed to only have a single fragment in every ring. > > This wouldn't solve the problem of pathologically badly fragmented, > > or very long URB, but there are simpler solutions to that. > > > > I believe that very long bulk transfers can be split into > > separate TD on USB block boundaries - so they can be safely > > added to the ring in several pieces. > > No, that's not a good idea. Think about the case where the device > driver sets up an IN transfer to receive a specific amount of data from > the device. The device can choose to send less data, resulting in a > short packet TD completion event. IN transfers are unlikely to be badly fragmented. The problem arise with OUT transfers. > Frankly, since these are your first kernel patches, I can't trust any > big sweeping changes you make unless I can throughly understand them. > That's just a fact about being a kernel maintainer; I have to understand > the code that gets contributed because the contributor is highly likely > to disappear, leaving me to fix bugs in their code years later. > > So, I would recommend you work on small fixes and earn my trust before > making big sweeping changes. Let's get the scatter gather issue fixed > before you completely re-architect the xHCI driver, okay? :) I've only been thinking about those changes. Although I'm new to the usb mailing list, I've been writing Unix device drivers, comms stack drivers and low level hardware interfacing code for most of the last 30 years. David -- To unsubscribe from this list: send the line "unsubscribe linux-usb" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
