DPDK patches and discussions
 help / color / mirror / Atom feed
From: John Daley <johndale@cisco.com>
To: ferruh.yigit@intel.com, arybchenko@solarflare.com
Cc: dev@dpdk.org, John Daley <johndale@cisco.com>,
	Hyong Youb Kim <hyonkim@cisco.com>
Subject: [dpdk-dev] [PATCH 4/5] net/enic: support flow API RSS ranges on outer headers
Date: Tue, 14 Apr 2020 18:06:40 -0700	[thread overview]
Message-ID: <20200415010641.5195-4-johndale@cisco.com> (raw)
In-Reply-To: <20200415010641.5195-1-johndale@cisco.com>

Support rte_flow RSS action on outer headers (level 0). RSS ranges on
the non-default port is OK.
Restrictions:
 - The RETA is ignored. The hash function is simply applied across
   the RSS queue range.
 - The queues used in the RSS group must be sequential.
 - There is a performance hit if the number of queues is not a power
   of 2.

Signed-off-by: John Daley <johndale@cisco.com>
Reviewed-by: Hyong Youb Kim <hyonkim@cisco.com>
---
 doc/guides/nics/enic.rst        |  4 +++
 drivers/net/enic/enic_fm_flow.c | 56 ++++++++++++++++++++++-----------
 2 files changed, 41 insertions(+), 19 deletions(-)

diff --git a/doc/guides/nics/enic.rst b/doc/guides/nics/enic.rst
index 65e536d422..aa4fdc0e39 100644
--- a/doc/guides/nics/enic.rst
+++ b/doc/guides/nics/enic.rst
@@ -459,6 +459,10 @@ PKT_RX_VLAN_STRIPPED mbuf flags would not be set. This mode is enabled with the
     packets and then receive them normally. These require 1400 series VIC adapters
     and latest firmware.
   - RAW items are limited to matching UDP tunnel headers like VXLAN.
+  - For 1400 VICs, all flows using the RSS action on a port use same hash
+    configuration. The RETA is ignored. The queues used in the RSS group must be
+    sequential. There is a performance hit if the number of queues is not a power of 2.
+    Only level 0 (outer header) RSS is allowed.
 
 - **Statistics**
 
diff --git a/drivers/net/enic/enic_fm_flow.c b/drivers/net/enic/enic_fm_flow.c
index 8d715fc436..86efeffc64 100644
--- a/drivers/net/enic/enic_fm_flow.c
+++ b/drivers/net/enic/enic_fm_flow.c
@@ -1174,8 +1174,8 @@ enic_fm_copy_action(struct enic_flowman *fm,
 				actions->conf;
 
 			/*
-			 * If other fate kind is set, fail.  Multiple
-			 * queue actions are ok.
+			 * If fate other than QUEUE or RSS, fail. Multiple
+			 * rss and queue actions are ok.
 			 */
 			if ((overlap & FATE) && first_rq)
 				goto unsupported;
@@ -1185,6 +1185,7 @@ enic_fm_copy_action(struct enic_flowman *fm,
 			fm_op.fa_op = FMOP_RQ_STEER;
 			fm_op.rq_steer.rq_index =
 				enic_rte_rq_idx_to_sop_idx(queue->index);
+			fm_op.rq_steer.rq_count = 1;
 			fm_op.rq_steer.vnic_handle = vnic_h;
 			ret = enic_fm_append_action_op(fm, &fm_op, error);
 			if (ret)
@@ -1219,27 +1220,44 @@ enic_fm_copy_action(struct enic_flowman *fm,
 			uint16_t i;
 
 			/*
-			 * Hardware does not support general RSS actions, but
-			 * we can still support the dummy one that is used to
-			 * "receive normally".
+			 * If fate other than QUEUE or RSS, fail. Multiple
+			 * rss and queue actions are ok.
+			 */
+			if ((overlap & FATE) && first_rq)
+				goto unsupported;
+			first_rq = false;
+			overlap |= FATE;
+
+			/*
+			 * Hardware only supports RSS actions on outer level
+			 * with default type and function. Queues must be
+			 * sequential.
 			 */
 			allow = rss->func == RTE_ETH_HASH_FUNCTION_DEFAULT &&
-				rss->level == 0 &&
-				(rss->types == 0 ||
-				 rss->types == enic->rss_hf) &&
-				rss->queue_num == enic->rq_count &&
-				rss->key_len == 0;
-			/* Identity queue map is ok */
-			for (i = 0; i < rss->queue_num; i++)
-				allow = allow && (i == rss->queue[i]);
+				rss->level == 0 && (rss->types == 0 ||
+				rss->types == enic->rss_hf) &&
+				rss->queue_num <= enic->rq_count &&
+				rss->queue[rss->queue_num - 1] < enic->rq_count;
+
+
+			/* Identity queue map needs to be sequential */
+			for (i = 1; i < rss->queue_num; i++)
+				allow = allow && (rss->queue[i] ==
+					rss->queue[i - 1] + 1);
 			if (!allow)
 				goto unsupported;
-			if (overlap & FATE)
-				goto unsupported;
-			/* Need MARK or FLAG */
-			if (!(overlap & MARK))
-				goto unsupported;
-			overlap |= FATE;
+
+			memset(&fm_op, 0, sizeof(fm_op));
+			fm_op.fa_op = FMOP_RQ_STEER;
+			fm_op.rq_steer.rq_index =
+				enic_rte_rq_idx_to_sop_idx(rss->queue[0]);
+			fm_op.rq_steer.rq_count = rss->queue_num;
+			fm_op.rq_steer.vnic_handle = vnic_h;
+			ret = enic_fm_append_action_op(fm, &fm_op, error);
+			if (ret)
+				return ret;
+			ENICPMD_LOG(DEBUG, "create QUEUE action rq: %u",
+				    fm_op.rq_steer.rq_index);
 			break;
 		}
 		case RTE_FLOW_ACTION_TYPE_PORT_ID: {
-- 
2.22.0


  parent reply	other threads:[~2020-04-15  1:08 UTC|newest]

Thread overview: 6+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-04-15  1:06 [dpdk-dev] [PATCH 1/5] net/enic: fix action reordering John Daley
2020-04-15  1:06 ` [dpdk-dev] [PATCH 2/5] net/enic: flow manager API update John Daley
2020-04-15  1:06 ` [dpdk-dev] [PATCH 3/5] net/enic: change Rx queue ordering to enable RSS action John Daley
2020-04-15  1:06 ` John Daley [this message]
2020-04-15  1:06 ` [dpdk-dev] [PATCH 5/5] net/enic: allow multiple mark and flag actions John Daley
2020-04-15 16:56 ` [dpdk-dev] [PATCH 1/5] net/enic: fix action reordering Ferruh Yigit

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20200415010641.5195-4-johndale@cisco.com \
    --to=johndale@cisco.com \
    --cc=arybchenko@solarflare.com \
    --cc=dev@dpdk.org \
    --cc=ferruh.yigit@intel.com \
    --cc=hyonkim@cisco.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).