> -----Original Message-----
> From: Guo, Jia <jia....@intel.com>
> Sent: Wednesday, May 13, 2020 7:20 AM
> To: Zhang, Qi Z <qi.z.zh...@intel.com>; Xing, Beilei <beilei.x...@intel.com>;
> Wu, Jingjing <jingjing...@intel.com>
> Cc: Ye, Xiaolong <xiaolong...@intel.com>; dev@dpdk.org
> Subject: Re: [dpdk-dev] net/iavf: fix rss algorithm configure issue
>
> hi, qi
>
> On 5/12/2020 10:21 AM, Zhang, Qi Z wrote:
> >
> >> -----Original Message-----
> >> From: Zhang, Qi Z
> >> Sent: Tuesday, May 12, 2020 10:18 AM
> >> To: Guo, Jia <jia....@intel.com>; Xing, Beilei
> >> <beilei.x...@intel.com>; Wu, Jingjing <jingjing...@intel.com>
> >> Cc: Ye, Xiaolong <xiaolong...@intel.com>; dev@dpdk.org
> >> Subject: RE: [dpdk-dev] net/iavf: fix rss algorithm configure issue
> >>
> >>
> >>
> >>> -----Original Message-----
> >>> From: Guo, Jia <jia....@intel.com>
> >>> Sent: Sunday, May 10, 2020 4:28 AM
> >>> To: Xing, Beilei <beilei.x...@intel.com>; Zhang, Qi Z
> >>> <qi.z.zh...@intel.com>; Wu, Jingjing <jingjing...@intel.com>
> >>> Cc: Ye, Xiaolong <xiaolong...@intel.com>; dev@dpdk.org; Guo, Jia
> >>> <jia....@intel.com>
> >>> Subject: [dpdk-dev] net/iavf: fix rss algorithm configure issue
> >>>
> >>> When configure rss rule, the etherdev rss hash function type should
> >>> be mapped to the corresponding virtchnl rss algorithm type.
> >>>
> >>> Fixes: 7be10c3004be ("net/iavf: add RSS configuration for VF")
> >>> Signed-off-by: Jeff Guo <jia....@intel.com>
> >>> ---
> >>> drivers/net/iavf/iavf_hash.c | 20 +++++++++++---------
> >>> 1 file changed, 11 insertions(+), 9 deletions(-)
> >>>
> >>> diff --git a/drivers/net/iavf/iavf_hash.c
> >>> b/drivers/net/iavf/iavf_hash.c index 6359ead76..173b7633f 100644
> >>> --- a/drivers/net/iavf/iavf_hash.c
> >>> +++ b/drivers/net/iavf/iavf_hash.c
> >>> @@ -47,7 +47,7 @@ struct iavf_hash_match_type {
> >>>
> >>> struct iavf_rss_meta {
> >>> struct virtchnl_proto_hdrs *proto_hdrs;
> >>> - uint32_t hash_function;
> >>> + enum virtchnl_rss_algorithm rss_algorithm;
> >>> };
> >>>
> >>> struct iavf_hash_flow_cfg {
> >>> @@ -1001,13 +1001,15 @@ iavf_hash_parse_action(struct
> >>> iavf_pattern_match_item *pattern_match_item,
> >>>
> >>> /* Check hash function and save it to rss_meta.
> >>> */
> >>> if (rss->func ==
> RTE_ETH_HASH_FUNCTION_SIMPLE_XOR)
> >>> - rss_meta->hash_function =
> >>> - RTE_ETH_HASH_FUNCTION_SIMPLE_XOR;
> >>> -
> >>> - if (rss->func ==
> >>> - RTE_ETH_HASH_FUNCTION_SYMMETRIC_TOEPLITZ)
> >>> - rss_meta->hash_function =
> >>> - RTE_ETH_HASH_FUNCTION_SYMMETRIC_TOEPLITZ;
> >>> + rss_meta->rss_algorithm =
> >>> + VIRTCHNL_RSS_ALG_R_ASYMMETRIC;
> >> RTE_ETH_HASH_FUNCTION_SYMMETRIC_TOEPLITZ -->
> >> VIRTCHNL_RSS_ALG_R_ASYMMETRIC
> > A typo, it is
> > RTE_ETH_HASH_FUNCTION_SIMPLE_XOR -->
> VIRTCHNL_RSS_ALG_R_ASYMMETRIC which looks not correct.
> >> Is it should be VIRTCHNL_RSS_ALG_XOR_SYMMETRIC?
> >>
> >>
>
> At currently driver, VIRTCHNL_RSS_ALG_R_ASYMMETRIC is used to
> identify simple XOR.
I didn't get it..
After apply your patch, we have below code which does not make sense,
/* Check hash function and save it to rss_meta. */
if (rss->func == RTE_ETH_HASH_FUNCTION_SIMPLE_XOR)
rss_meta->rss_algorithm =
VIRTCHNL_RSS_ALG_R_ASYMMETRIC;
else if...
ASYMMETRIC is not kind of symmetric hash and should not be mapped to simple xor.
>
>
> >>> + else if (rss->func ==
> >>> +
> RTE_ETH_HASH_FUNCTION_SYMMETRIC_TOEPLITZ)
> >>> + rss_meta->rss_algorithm =
> >>> + VIRTCHNL_RSS_ALG_TOEPLITZ_SYMMETRIC;
> >>> + else
> >>> + rss_meta->rss_algorithm =
> >>> + VIRTCHNL_RSS_ALG_TOEPLITZ_ASYMMETRIC;
> >>>
> >>> type_match_item =
> >>> rte_zmalloc("iavf_type_match_item",
> >>> @@ -1126,7 +1128,7 @@ iavf_hash_create(__rte_unused struct
> >>> iavf_adapter *ad,
> >>> }
> >>>
> >>> rss_cfg->proto_hdrs = *rss_meta->proto_hdrs;
> >>> - rss_cfg->rss_algorithm = rss_meta->hash_function;
> >>> + rss_cfg->rss_algorithm = rss_meta->rss_algorithm;
> >>>
> >>> ret = iavf_add_del_rss_cfg(ad, rss_cfg, true);
> >>> if (!ret) {
> >>> --
> >>> 2.20.1
