[PATCH 3/3] net: hisilicon: new hip04 ethernet driver

[Eric Dumazet]

On Tue, 2014-03-25 at 18:05 +0100, Arnd Bergmann wrote:
> On Tuesday 25 March 2014 10:00:30 Florian Fainelli wrote:
> > 2014-03-25 1:12 GMT-07:00 Arnd Bergmann <arnd at arndb.de>:
> > > On Tuesday 25 March 2014 12:06:31 Zhangfei Gao wrote:
> > >> Dear Arnd
> > >>
> > >> On Mon, Mar 24, 2014 at 11:18 PM, Arnd Bergmann <arnd at arndb.de> wrote:
> > >> > On Monday 24 March 2014 22:14:56 Zhangfei Gao wrote:
> > >> >
...
> > >> > I think you still need to find a solution to ensure that the tx reclaim is
> > >> > called eventually through a method other than start_xmit.
> > >>
> > >> In the iperf stress test, if move reclaim to poll, there is some
> > >> error, sometimes sending zero packets.
> > >> While keep reclaim in the xmit to reclaim transmitted packets looks
> > >> stable in the test,
> > >> There TX_DESC_NUM desc can be used.
> > >
> > > What I meant is that you need a correct implementation, presumably
> > > you added a bug when you moved the function to poll(), and also you
> > > forgot to add a timer.
> > 
> > Using a timer to ensure completion of TX packets is a trick that
> > worked in the past, but now that the networking stack got smarter,
> > this might artificially increase the processing time of packets in the
> > transmit path, and this will defeat features like TCP small queues
> > etc.. as could be seen with the mvneta driver [1]. The best way really
> > is to rely on TX completion interrupts when those exist as they cannot
> > lie about the hardware status (in theory) and they should provide the
> > fastest way to complete TX packets.
> 
> By as Zhangfei Gao pointed out, this hardware does not have a working
> TX completion interrupt. Using timers to do this has always just been
> a workaround for broken hardware IMHO.

For this kind of drivers, calling skb_orphan() from ndo_start_xmit() is
mandatory.