SSH is usually pretty good at it, but dropping the tx dmamap segments helps a lot as Kim said. On 2 Apr 2015 9:45 am, "Kimberley Manning" <[email protected]> wrote:
> On 02/04/15 07:51, Stefan Fritsch wrote: > >> On Wednesday 01 April 2015 12:46:47, Kimberley Manning wrote: >> >>> This diff makes the pcn driver use m_defrag for fragmented mbuf >>>>> chains,> >>>>> >>>> >>>> I like this kind of cleanups. As for vio(4) could you try the >>>> diff or are you looking for testers? >>>> >>> >>> Thanks. I've tried both the vio and pcn diffs and they worked for >>> me, so dlg told me to write here. I'm not sure what happens to >>> diffs after this point to be honest. >>> >> >> I can take a look at the vio diff and commit it in a few days. >> >> BTW, is there a way to trigger heavily fragmented mbufs in order to >> test the code path? >> > > Reducing the tx max segments count worked for me (dlg's suggestion). > >
