On Tue, Sep 18, 2012 at 2:34 PM, Javier Marcet <jmarcet@xxxxxxxxx> wrote: > I can't say how happy I am that you were wrong in your guess. Quoting > Konrad Rzeszutek Wilk: Well, you haven't proven me wrong yet. :-) > """ > The issue as I understand is that the DVB drivers allocate their > buffers from 0->4GB most (all the time?) so they never have to do > bounce-buffering. > > While the pv-ops one ends up quite frequently doing the > bounce-buffering, which implies that the DVB drivers end up allocating > their buffers above the 4GB. > This means we end up spending some CPU time (in the guest) copying the > memory from >4GB to 0-4GB region (And vice-versa). > """ > > You can see the original thread where this was found, together with a > working patch, here: > > http://lists.xen.org/archives/html/xen-devel/2012-01/msg01927.html As far as I can read, the patch has never been confirmed to work. The user mentioned upgrading to an updated kernel and seeing a slight decrease in load: http://lists.xen.org/archives/html/xen-devel/2012-01/msg02166.html Further, the reason many of the drivers in question require the memory to be in the 0-4GB memory region is due to some hardware not being able to DMA to memory > 4GB. Such a change would have to be tested with every board that does scatter/gather, and the framework would likely have to change to explicitly allow the board driver to specify whether it supports memory > 4GB. In short, this is a useful bit of information, but not clear whether it would actually solve the underlying problem. Again, I would be happy to be proven wrong, but there appears to still be quite a bit of work required for such. I would suggest trying the patch yourself to see if it has any visible effect on the problem. Devin -- Devin J. Heitmueller - Kernel Labs http://www.kernellabs.com -- To unsubscribe from this list: send the line "unsubscribe linux-media" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
