RE: [EXT] Re: [PATCH v2 RESEND 1/1] net: fec: add xdp and page pool statistics

[Shenwei Wang]

> -----Original Message-----
> From: Alexander Lobakin <alexandr.lobakin@xxxxxxxxx>
> Sent: Monday, November 14, 2022 7:57 AM
> To: Andrew Lunn <andrew@xxxxxxx>
> Cc: Alexander Lobakin <alexandr.lobakin@xxxxxxxxx>; Shenwei Wang
> <shenwei.wang@xxxxxxx>; Paolo Abeni <pabeni@xxxxxxxxxx>; David S. Miller
> <davem@xxxxxxxxxxxxx>; Eric Dumazet <edumazet@xxxxxxxxxx>; Jakub
> Kicinski <kuba@xxxxxxxxxx>; Alexei Starovoitov <ast@xxxxxxxxxx>; Daniel
> Borkmann <daniel@xxxxxxxxxxxxx>; Jesper Dangaard Brouer
> <hawk@xxxxxxxxxx>; John Fastabend <john.fastabend@xxxxxxxxx>;
> netdev@xxxxxxxxxxxxxxx; linux-kernel@xxxxxxxxxxxxxxx; imx@xxxxxxxxxxxxxxx
> Subject: [EXT] Re: [PATCH v2 RESEND 1/1] net: fec: add xdp and page pool
> statistics
> 
> Caution: EXT Email
> 
> From: Andrew Lunn <andrew@xxxxxxx>
> Date: Mon, 14 Nov 2022 14:50:54 +0100
> 
> > >    What is your machine and how fast your link is?
> >
> > Some FEC implementations are Fast Ethernet. Others are 1G.
> >
> > I expect Shenwei is testing on a fast 64 bit machine with 1G, but
> > there are slow 32bit machines with Fast ethernet or 1G.
> 
> Okay. I can say I have link speed on 1G on MIPS32, even on those which are 1-
> core 600 MHz. And when I was adding more driver stats, all based on
> u64_stats_t, I couldn't spot any visible regression.
> 100-150 Kbps maybe?
> 

I did implement a quick version of u64_stats_t counters, and the performance impact
was about 1~3Mbps on the i.MX8QXP which is 1.2GHz ARM64 Dual Core platform, which
is about 1.5% performance decrease.

Regards,
Shenwei

> >
> >      Andrew
> 
> Thanks,
> Olek