[dpdk-dev] [PATCH 2/4] i40e: split function for input set change of hash and fdir
jingjing.wu at intel.com
Thu Jan 21 02:28:59 CET 2016
Thanks for your comments. You are correct, I removed the I40E_INSET_FLEX_PAYLOAD
from valid fdir input set values, and this is one reason why I splited function for input set
change of hash and and it is because all flex payload configuration can be
set in struct rte_fdir_conf during device configure phase. And it is a more flexible configuration
including flexpayload's selection, input set selection by word and mask setting in bits.
If I enable it in the input set change API, it will be duplicate. And the input set change on
flexible payload only on word, just some ability compared with rte_fdir_conf.
If flexible selection isn't done in struct rte_fdir_conf, the input set
selection in input set change API doesn't make sense. If flexible selection is done in
struct rte_fdir_conf, why not selection input set in struct rte_fdir_conf at the same time?
And about you concern, "when application has to run on an old NIC and on a new one",
The rte_fdir_conf is for each eth_dev, so it will be fine.
> -----Original Message-----
> From: Chilikin, Andrey
> Sent: Thursday, January 21, 2016 4:05 AM
> To: Wu, Jingjing; dev at dpdk.org
> Cc: Zhang, Helin; Pei, Yulong; Ananyev, Konstantin
> Subject: RE: [PATCH 2/4] i40e: split function for input set change of hash and
> Hi Jingjing,
> As I can see this patch not only splits fdir functionality from common
> fdir/hash code but also removes compatibility with DPDK 2.2 as it deletes
> I40E_INSET_FLEX_PAYLOAD from valid fdir input set values. Yes, flexible
> payload configuration can be set for fdir separately at the port initialization,
> but this is more legacy from the previous generations of NICs which did not
> support dynamic input set configuration. I believe it would better to have
> I40E_INSET_FLEX_PAYLOAD valid for fdir input set same as in DPDK 2.2. So in
> legacy mode, when application has to run on an old NIC and on a new one,
> only legacy configuration would be used, but for applications targeting new
> HW single point of configuration would be used instead of mix of two.
More information about the dev