On Fri, 2016-01-29 at 11:08 -0800, David Miller wrote: > From: Eric Dumazet <eric.duma...@gmail.com> > Date: Fri, 29 Jan 2016 10:40:10 -0800 > > > I would try following ixgbe fix (sorry, totally untested, but you get > > the idea...) > > > > diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c > > b/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c > > index c4003a88bbf6..7ba64ed463a6 100644 > > --- a/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c > > +++ b/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c > > @@ -1525,7 +1525,7 @@ static bool ixgbe_alloc_mapped_page(struct ixgbe_ring > > *rx_ring, > > > > bi->dma = dma; > > bi->page = page; > > - bi->page_offset = 0; > > + bi->page_offset = NET_IP_ALIGN; > > > > return true; > > } > > Hmmm, wouldn't this waste 2 bytes on all pages? Or just the head one with > the headers?
On arches where NET_IP_ALIGN = 2, where it makes sense to get a proper alignement, even for the copy of headers into skb->head, with same source/destination alignment. We 'waste' already 66 bytes for typical ethernet+IPv4+TCP headers. Lets remind we allocate half a page (2048 bytes) for an ethernet frame anyway. We have plenty of room.
