RE: [PATCH 6/7] mlx4_en: Adding rxhash support

[Eric Dumazet <eric.dumazet@gmail.com>]

Le mardi 18 octobre 2011 à 07:36 +0000, Yevgeny Petrilin a écrit :
> > >  	rss_context->flags = rss_mask;
> > > +	rss_context->hash_fn = 1;
> > > +	for (i = 0; i < 10; i++)
> > > +		rss_context->rss_key[i] = random32();
> > >
> > 
> > Thats bit of a problem : Two NICS will have different seeds, and thus provide different rxhash for a given flow. A bonding of two NICS will
> > not be able to provide a consistent rxhash.
> > 
> > drivers/net/ethernet/intel/igb/igb_main.c uses a static table to avoid this problem.
> > 
> 
> Hello Eric, thanks for your review.
> 
> I agree that in this case two ports will have different seeds.
> But even if we use static values for the key, what about bonding of 2 NICs from different vendors?
> How can we ensure we get same rxhash value for all NICs?
> 
> There are also other drivers that use random values as well, for example:
> drivers/net/ethernet/broadcom/bnx2x/bnx2x_cmn.c
> 

What is the gain using random values ?

Usually, we tend to have same hardware in a single machine, or we use
active-backup bonding mode, and an active slave flip can change rxhash
values with litle effect, since this happens not often.

I really prefer not random values, because it allows to have replayable
configurations : For a given tcp flow, the same rxhash value is given
and same cpu target in RPS. Its way easier to tune your machine for some
workloads.