On Thu, 14 Jul 2005, Pete Zaitcev wrote: > On Thu, 14 Jul 2005 10:22:59 -0400 (EDT), Alan Stern <[EMAIL PROTECTED]> > wrote: > > > However this ignores the matter of retrying submissions that fail with > > -ENXIO, -EAGAIN, or -ENOMEM. Maybe you could add an additional argument > > to usb_sg_submit to specify whether such failures should be retried. > > I don't like such functions with variable behaviour and the duplication > is not big enough here.
Okay. > As for usb_sg_cancel in case of an error, the logic was that it makes > no difference, so why do it? It makes no difference in terms of the values you return to the caller, but it does make a difference in terms of performance. It's better to fail-fast than to fail-slow. On the other hand, it's true that this is a low-probability failure mode and hence the performance isn't very important. > Even if I call it, I cannot return with > anything but zero error code (unless the very first submit fails). That's right. > But calling this cancel adds an opportunity for racing somewhere. No it doesn't. The opportunity was there all along. Remember, the caller could call usb_sg_cancel from another thread at precisely this time. > I do not trust our code in this area. Remember, ub must not crash. > Unlinking is the crashiest thing in the whole stack, so heck no. I can understand that at the moment you don't trust the new usb_sg code. Without a lot of checking and testing, you shouldn't. However it's not _impossibly_ difficult, and with some work you ought to be able to resolve all the possible races. (With the changes I suggested it should be pretty close already.) Trusting unlinks is a separate matter. However I think they are pretty well under control now. I can't recall hearing of any problems in the HCD code relating to buggy unlinks for quite some time. Alan Stern ------------------------------------------------------- SF.Net email is sponsored by: Discover Easy Linux Migration Strategies from IBM. Find simple to follow Roadmaps, straightforward articles, informative Webcasts and more! Get everything you need to get up to speed, fast. http://ads.osdn.com/?ad_id=7477&alloc_id=16492&op=click _______________________________________________ linux-usb-devel@lists.sourceforge.net To unsubscribe, use the last form field at: https://lists.sourceforge.net/lists/listinfo/linux-usb-devel
