Re: [Xen-devel] XEN_NETIF_NR_SLOTS_MIN

["Jan Beulich" <JBeulich@xxxxxxxx>]

>>> On 07.09.15 at 17:58, <wei.liu2@xxxxxxxxxx> wrote:
> On Mon, Sep 07, 2015 at 09:15:40AM -0600, Jan Beulich wrote:
>> >>> On 07.09.15 at 16:56, <wei.liu2@xxxxxxxxxx> wrote:
>> > Linux netfront is able to consolidate several slots into reasonable
>> > number of frags and make sure the number of frags won't exceed limit.
>> > See xennet_fill_frags.
>> 
>> That's too late - xennet_get_responses() would have rejected such
>> a packet already afaict (error message "Too many slots").
>> 
> 
> Right, that's a bit unfortunate.  We might need to have that on the
> protocol level, say, backend should not push more than X slots to
> frontend.
> 
> OOI are you seeing this in real world?

Yes, with our netback (which I'm only now getting an equivalent of
David's coalescing patch added) I could easily see these "Too many
frags" (in our netfront, still using the old naming) messages. Clearly
the patch to netback will hide the errors, but it still shows that
netfront doesn't cope with old netback (and that's even when the
netfront used sees the same smaller MAX_SKB_FRAGS that netback
also uses, i.e. the situation would be worse with a netback on an
older kernel with the one larger MAX_SKB_FRAGS).

Jan


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel