RE: [PATCH] xen/netback: calculate correctly the SKB slots.

[Ian Campbell <Ian.Campbell@citrix.com>]

On Tue, 2012-05-22 at 20:01 +0100, Simon Graham wrote:
> > >
> > > > >  	int i, copy_off;
> > > > >
> > > > >  	count = DIV_ROUND_UP(
> > > > > -			offset_in_page(skb->data)+skb_headlen(skb),
> > PAGE_SIZE);
> > > > > +			offset_in_page(skb->data + skb_headlen(skb)),
> > PAGE_SIZE);
> > > >
> > > > The new version would be equivalent to:
> > > > 	count = offset_in_page(skb->data + skb_headlen(skb)) != 0;
> > > > which is not right, as netbk_gop_skb() will use one slot per page.
> > >
> > > Just outside the context of this patch we separately count the frag
> > > pages.
> > >
> > > However I think you are right if skb->data covers > 1 page, since the
> > > new version can only ever return 0 or 1. I expect this patch papers
> > over
> > > the underlying issue by not stopping often enough, rather than
> > actually
> > > fixing the underlying issue.
> > 
> > Ah, any thoughts? Have you guys seen this behavior as well?
> 
> We ran into this same problem and the fix we've been running with for
> a while now (been meaning to submit it!) is:
> 
> diff --git a/drivers/net/xen-netback/netback.c b/drivers/net/xen-netback/netback.c
> index c2669b8..7925bd3 100644
> --- a/drivers/net/xen-netback/netback.c
> +++ b/drivers/net/xen-netback/netback.c
> @@ -312,8 +312,7 @@ unsigned int xen_netbk_count_skb_slots(struct xenvif *vif, struct sk_buff *skb)
>         unsigned int count;
>         int i, copy_off;
> 
> -       count = DIV_ROUND_UP(
> -                       offset_in_page(skb->data)+skb_headlen(skb), PAGE_SIZE);
> +       count = DIV_ROUND_UP(skb_headlen(skb), PAGE_SIZE);
> 
>         copy_off = skb_headlen(skb) % PAGE_SIZE;
> 
> The rationale for this is that if the header spanned a page boundary,
> you would calculate that it needs 2 slots for the header BUT
> netback_gop_skb copies the header into the start of the page so only
> needs one slot (and only decrements the count of inuse entries by 1).

That sounds very plausible indeed!

Please can format this as a commit message and resend with a
Signed-off-by.

many thanks,
Ian.

> 
> We found this running with a VIF bridged to a USB 3G Modem where
> skb->data started near the end of a page so the header would always
> span the page boundary.
> 
> It was very easy to get the VIF to stop processing frames with the old
> code and we have not seen any problems since applying this patch.
> 
> Simon
>