[ovs-dev] [v3] dpif/dpcls: limit count subtable search info logs

[kumar Amber]

From: Harry van Haaren <harry.van.haaren at intel.com>

This commit avoids many instances of "using subtable X for miniflow (x,y)"
in the ovs-vswitchd log when using the DPCLS Autovalidator. This occurs
when no specialized subtable is found, and the generic "_any" version of
the avx512 subtable search implementation was used. This change logs the
subtable usage once, avoiding duplicates.

Signed-off-by: Harry van Haaren <harry.van.haaren at intel.com>

---
v3:
- add comments from Flavio
- add documentation update
---
 Documentation/topics/dpdk/bridge.rst   | 31 ++++++++++++++++++++++++++
 lib/dpif-netdev-lookup-avx512-gather.c |  4 ++--
 2 files changed, 33 insertions(+), 2 deletions(-)

diff --git a/Documentation/topics/dpdk/bridge.rst b/Documentation/topics/dpdk/bridge.rst
index 0f70a0cad..e74c839b5 100644
--- a/Documentation/topics/dpdk/bridge.rst
+++ b/Documentation/topics/dpdk/bridge.rst
@@ -182,6 +182,37 @@ chosen, and the 2nd occurance of that priority is not used. Put in logical
 terms, a subtable is chosen if its priority is greater than the previous
 best candidate.
 
+Optimizing Specific Subtable Search
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+The DPCLS wildcarding engine can be "specialized" to handle specific subtable
+searching even faster than the generic subtable search implementation.Not all
+subtable searches are specialized, if you see a message like the following in
+your OVS output, please run the below commands to inform OVS community of the
+subtables in use in your deployment, and we can optimize these subtables for
+you in the next release.
+
+"Using non-specialized AVX512 lookup for subtable (7,2) and possibly others."
+
+(Note that the numbers 7 and 2 will likely be different in your logs, these
+are the numbers OVS community requires to specialize the subtable in your
+deployment!)
+
+If you see this log message, please run the command
+"ovs-appctl dpctl/dump-flows -m", which results in output like this:
+
+    ufid:82770b5d-ca38-44ff-8283-74ba36bd1ca5, skb_priority(0/0),skb_mark(0/0)
+    ,ct_state(0/0),ct_zone(0/0),ct_mark(0/0),ct_label(0/0),recirc_id(0),
+    dp_hash(0/0),in_port(pcap0),packet_type(ns=0,id=0),eth(src=00:00:00:00:00:
+    00/00:00:00:00:00:00,dst=ff:ff:ff:ff:ff:ff/00:00:00:00:00:00),eth_type(
+    0x8100),vlan(vid=1,pcp=0),encap(eth_type(0x0800),ipv4(src=127.0.0.1/0.0.0.0
+    ,dst=127.0.0.1/0.0.0.0,proto=17/0,tos=0/0,ttl=64/0,frag=no),udp(src=53/0,
+    dst=53/0)), packets:77072681, bytes:3545343326, used:0.000s, dp:ovs,
+    actions:vhostuserclient0, dp-extra-info:miniflow_bits(4,1)
+
+Please send an email to the OVS mailing list ovs-dev at openvswitch.org with
+the output of the "dp-extra-info:miniflow_bits(7,2)" values.
+
 CPU ISA Testing and Validation
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 
diff --git a/lib/dpif-netdev-lookup-avx512-gather.c b/lib/dpif-netdev-lookup-avx512-gather.c
index bc359dc4a..ced846aa7 100644
--- a/lib/dpif-netdev-lookup-avx512-gather.c
+++ b/lib/dpif-netdev-lookup-avx512-gather.c
@@ -411,8 +411,8 @@ dpcls_subtable_avx512_gather_probe(uint32_t u0_bits, uint32_t u1_bits)
      */
     if (!f && (u0_bits + u1_bits) < (NUM_U64_IN_ZMM_REG * 2)) {
         f = dpcls_avx512_gather_mf_any;
-        VLOG_INFO("Using avx512_gather_mf_any for subtable (%d,%d)\n",
-                  u0_bits, u1_bits);
+        VLOG_INFO_ONCE("Using non-specialized AVX512 lookup for subtable"
+                       " (%d,%d) and possibly others.", u0_bits, u1_bits);
     }
 
     return f;
-- 
2.25.1