Re: [dpdk-dev] [PATCH 3/3] net/i40e: configure RSS hash from RSS action

["Iremonger, Bernard" <bernard.iremonger@intel.com>]

Hi Chenxu,

> -----Original Message-----
> From: Di, ChenxuX <chenxux.di@intel.com>
> Sent: Monday, March 23, 2020 9:19 AM
> To: Iremonger, Bernard <bernard.iremonger@intel.com>; dev@dpdk.org;
> Xing, Beilei <beilei.xing@intel.com>; Zhang, Qi Z <qi.z.zhang@intel.com>;
> Doherty, Declan <declan.doherty@intel.com>
> Cc: Ananyev, Konstantin <konstantin.ananyev@intel.com>; Iremonger,
> Bernard <bernard.iremonger@intel.com>
> Subject: RE: [dpdk-dev] [PATCH 3/3] net/i40e: configure RSS hash from RSS
> action
> 
> Hi, Bernard
> This patch conflicts with the
> patch(http://patchwork.dpdk.org/project/dpdk/list/?series=9005). In 20.05
> release, there will be some changes for creating RSS flows to support more
> features and make it more flexible.
> Changes for RSS flow:
> - Configure the hash input set according to the packet type.
> - Configure hash function.
> - User can create more than one RSS flows and won't affect other rules when
> destory.
> For example, if you want to enable rss for ipv4-tcp, the command looks like
> this:
> flow create 0 ingress pattern eth / ipv4 / tcp / end actions rss types ipv4-tcp
> end queues end / end
> 
> So, perhaps you need rework your feature depending on that patch.
> 
> > -----Original Message-----
> > From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Bernard
> Iremonger
> > Sent: Wednesday, March 18, 2020 8:38 PM
> > To: dev@dpdk.org; Xing, Beilei <beilei.xing@intel.com>; Zhang, Qi Z
> > <qi.z.zhang@intel.com>; Doherty, Declan <declan.doherty@intel.com>
> > Cc: Ananyev, Konstantin <konstantin.ananyev@intel.com>; Iremonger,
> > Bernard <bernard.iremonger@intel.com>
> > Subject: [dpdk-dev] [PATCH 3/3] net/i40e: configure RSS hash from RSS
> > action
> >
> > handle RSS offload types in RSS action handle ipv4, ipv6 and udp RSS
> > patterns handle queue_num 0 in RSS action
> >
> > Signed-off-by: Bernard Iremonger <bernard.iremonger@intel.com>
> > ---
> >  drivers/net/i40e/i40e_flow.c | 46
> > ++++++++++++++++++++++++++++++++-------
> > -----
> >  1 file changed, 34 insertions(+), 12 deletions(-)
> >
> > diff --git a/drivers/net/i40e/i40e_flow.c
> > b/drivers/net/i40e/i40e_flow.c index d877ac2..b9a928c 100644
> > --- a/drivers/net/i40e/i40e_flow.c
> > +++ b/drivers/net/i40e/i40e_flow.c
> > @@ -4461,6 +4461,12 @@ i40e_flow_parse_rss_pattern(__rte_unused
> struct
> > rte_eth_dev *dev,  case RTE_FLOW_ITEM_TYPE_ETH:
> >  *action_flag = 1;
> >  break;
> > +case RTE_FLOW_ITEM_TYPE_IPV4:
> > +break;
> > +case RTE_FLOW_ITEM_TYPE_IPV6:
> > +break;
> > +case RTE_FLOW_ITEM_TYPE_UDP:
> > +break;
> >  case RTE_FLOW_ITEM_TYPE_VLAN:
> >  vlan_spec = item->spec;
> >  vlan_mask = item->mask;
> > @@ -4516,12 +4522,14 @@ i40e_flow_parse_rss_action(struct
> rte_eth_dev
> > *dev,  struct i40e_rte_flow_rss_conf *rss_config =  &filter->rss_conf;
> > struct i40e_rte_flow_rss_conf *rss_info = &pf->rss_info;
> > +struct rte_eth_rss_conf eth_rss_conf; int ret;
> >  uint16_t i, j, n, tmp;
> >  uint32_t index = 0;
> >  uint64_t hf_bit = 1;
> > +uint64_t rss_hf;
> >
> >  NEXT_ITEM_OF_ACTION(act, actions, index); -rss = act->conf;
> >
> >  /**
> >   * rss only supports forwarding,
> > @@ -4535,6 +4543,23 @@ i40e_flow_parse_rss_action(struct rte_eth_dev
> > *dev,  return -rte_errno;  }
> >
> > +rss = act->conf;
> > +rss_hf = rss->types;
> > +
> > +if (rss->queue_num == 0) {
> > +memset(&eth_rss_conf, 0, sizeof(eth_rss_conf)); ret =
> > +rte_eth_dev_rss_hash_conf_get(dev->data->port_id,
> > +&eth_rss_conf);
> > +if (ret != 0)
> > +return ret;
> > +
> > +eth_rss_conf.rss_hf |= rss_hf;
> > +ret = rte_eth_dev_rss_hash_update(dev->data->port_id,
> > +&eth_rss_conf);
> > +if (ret != 0)
> > +return ret;
> > +}
> > +
> 
> I don't think it is a good idea that update rss hash enable status in the parse
> function.
> it will update status while send command  flow validate xxxxxx which can't be
> destoried.
> 
> >  if (action_flag) {
> >  for (n = 0; n < 64; n++) {
> >  if (rss->types & (hf_bit << n)) {
> > @@ -4552,7 +4577,7 @@ i40e_flow_parse_rss_action(struct rte_eth_dev
> *dev,
> >   * continuous sequence and also to be part of RSS
> >   * queue index for this port.
> >   */
> > -if (conf_info->queue_region_number) {
> > +if (conf_info->queue_region_number && rss->queue_num > 0) {
> >  for (i = 0; i < rss->queue_num; i++) {  for (j = 0; j <
> > rss_info->conf.queue_num; j++) {  if (rss->queue[i] ==
> > rss_info->conf.queue[j]) @@ -4579,7 +4604,8 @@
> > i40e_flow_parse_rss_action(struct rte_eth_dev *dev,  }
> >
> >  /* Parse queue region related parameters from configuration */ -for
> > (n = 0; n < conf_info->queue_region_number; n++) {
> > +for (n = 0; n < conf_info->queue_region_number &&
> > +rss->queue_num > 0; n++) {
> >  if (conf_info->region[n].user_priority_num ||
> >  conf_info->region[n].flowtype_num) {
> >  if (!((rte_is_power_of_2(rss->queue_num)) && @@ -
> > 4674,14 +4700,6 @@ i40e_flow_parse_rss_action(struct rte_eth_dev *dev,
> > if (rss_config->queue_region_conf)  return 0;
> >
> > -if (!rss || !rss->queue_num) {
> > -rte_flow_error_set(error, EINVAL,
> > -RTE_FLOW_ERROR_TYPE_ACTION,
> > -act,
> > -"no valid queues");
> > -return -rte_errno;
> > -}
> > -
> >  for (n = 0; n < rss->queue_num; n++) {  if (rss->queue[n] >=
> > dev->data->nb_rx_queues) {  rte_flow_error_set(error, EINVAL, @@
> > -4692,7 +4710,7 @@ i40e_flow_parse_rss_action(struct rte_eth_dev *dev,
> > }  }
> >
> > -if (rss_info->conf.queue_num) {
> > +if (rss_info->conf.queue_num && rss->queue_num) {
> >  rte_flow_error_set(error, EINVAL,
> >  RTE_FLOW_ERROR_TYPE_ACTION,
> >  act,
> > @@ -4709,6 +4727,10 @@ i40e_flow_parse_rss_action(struct rte_eth_dev
> > *dev,  return rte_flow_error_set  (error, ENOTSUP,
> > RTE_FLOW_ERROR_TYPE_ACTION, act,
> >   "a nonzero RSS encapsulation level is not supported");
> > +if (rss->types > ETH_RSS_ESP)
> > +return rte_flow_error_set
> > +(error, ENOTSUP, RTE_FLOW_ERROR_TYPE_ACTION,
> > act,
> > + "pctype greater than max allowed");
> 
> At the patch(http://patchwork.dpdk.org/project/dpdk/list/?series=9005),
> the hash input set will be enable.
> So rss->types may has the value ETH_RSS_L3_SRC_ONLY (1ULL << 63) and
> ETH_RSS_L3_DST_ONLY (1ULL << 62) and ETH_RSS_L4_SRC_ONLY (1ULL <<
> 61) and ETH_RSS_L4_DST_ONLY  (1ULL << 60).
> 
> >  if (rss->key_len && rss->key_len > RTE_DIM(rss_config->key))  return
> > rte_flow_error_set  (error, ENOTSUP, RTE_FLOW_ERROR_TYPE_ACTION,
> act,
> > --
> > 2.7.4
> 

Yes, I will need to rework this patch based on top of your patch.

Regards,

Bernard.