From: Tomasz Kulasek <tomaszx.kulasek@intel.com>
To: dev@dpdk.org
Cc: declan.doherty@intel.com
Subject: [dpdk-dev] [PATCH] net/bonding: fix link status check
Date: Wed, 29 Nov 2017 15:53:49 +0100 [thread overview]
Message-ID: <20171129145349.43448-1-tomaszx.kulasek@intel.com> (raw)
Some devices needs more time to initialize and bring interface up. When
link is down the link properties are not valid, e.g. link_speed is
reported as 0 and this is not a valid speed for slave as well as for whole
bonding.
During NIC (and bonding) initialization there's concurrency between
updating link status and adding slave to the bonding.
This patch:
- adds delay before configuring bonding (if link is down) to be sure that
link status of new slave is valid,
- propagates information about link status from first active slave
instead of first slave at all, to be sure that link speed is valid.
Fixes: 6abd94d72ab5 ("net/bonding: fix check slaves link properties")
Signed-off-by: Tomasz Kulasek <tomaszx.kulasek@intel.com>
---
drivers/net/bonding/rte_eth_bond_8023ad_private.h | 1 +
drivers/net/bonding/rte_eth_bond_api.c | 50 ++++++++++++++++-------
2 files changed, 37 insertions(+), 14 deletions(-)
diff --git a/drivers/net/bonding/rte_eth_bond_8023ad_private.h b/drivers/net/bonding/rte_eth_bond_8023ad_private.h
index 433c700..0a72beb 100644
--- a/drivers/net/bonding/rte_eth_bond_8023ad_private.h
+++ b/drivers/net/bonding/rte_eth_bond_8023ad_private.h
@@ -180,6 +180,7 @@ struct mode8023ad_private {
rte_eth_bond_8023ad_ext_slowrx_fn slowrx_cb;
uint8_t external_sm;
+ uint8_t slave_link_valid;
struct rte_eth_link slave_link;
/***< slave link properties */
diff --git a/drivers/net/bonding/rte_eth_bond_api.c b/drivers/net/bonding/rte_eth_bond_api.c
index 980e636..6b65fac 100644
--- a/drivers/net/bonding/rte_eth_bond_api.c
+++ b/drivers/net/bonding/rte_eth_bond_api.c
@@ -33,6 +33,7 @@
#include <string.h>
+#include <rte_cycles.h>
#include <rte_mbuf.h>
#include <rte_malloc.h>
#include <rte_ethdev.h>
@@ -239,6 +240,7 @@
struct bond_dev_private *internals;
struct rte_eth_link link_props;
struct rte_eth_dev_info dev_info;
+ int retries = 10;
bonded_eth_dev = &rte_eth_devices[bonded_port_id];
internals = bonded_eth_dev->data->dev_private;
@@ -262,6 +264,20 @@
return -1;
}
+ /* Some devices needs more time to initialize and bring interface up.
+ * While link status up is preferable we wait some time to be sure that
+ * link status of slave is valid.
+ */
+ if (slave_eth_dev->data->dev_link.link_status == ETH_LINK_DOWN) {
+ rte_delay_ms(100);
+ rte_eth_link_get_nowait(slave_port_id, &link_props);
+ while ((link_props.link_status == ETH_LINK_DOWN) && (retries > 0)) {
+ rte_delay_ms(100);
+ rte_eth_link_get_nowait(slave_port_id, &link_props);
+ retries--;
+ }
+ }
+
slave_add(internals, slave_eth_dev);
/* We need to store slaves reta_size to be able to synchronize RETA for all
@@ -270,15 +286,16 @@
internals->slaves[internals->slave_count].reta_size = dev_info.reta_size;
if (internals->slave_count < 1) {
+
+ /* Reset link status information for bonded slaves (it will be taken
+ * from first active slave) */
+ internals->mode4.slave_link_valid = 0;
+
/* if MAC is not user defined then use MAC of first slave add to
* bonded device */
if (!internals->user_defined_mac)
mac_address_set(bonded_eth_dev, slave_eth_dev->data->mac_addrs);
- /* Inherit eth dev link properties from first slave */
- link_properties_set(bonded_eth_dev,
- &(slave_eth_dev->data->dev_link));
-
/* Make primary slave */
internals->primary_port = slave_port_id;
internals->current_primary_port = slave_port_id;
@@ -302,14 +319,6 @@
internals->tx_offload_capa &= dev_info.tx_offload_capa;
internals->flow_type_rss_offloads &= dev_info.flow_type_rss_offloads;
- if (link_properties_valid(bonded_eth_dev,
- &slave_eth_dev->data->dev_link) != 0) {
- RTE_BOND_LOG(ERR, "Invalid link properties for slave %d"
- " in bonding mode %d", slave_port_id,
- internals->mode);
- return -1;
- }
-
/* RETA size is GCD of all slaves RETA sizes, so, if all sizes will be
* the power of 2, the lower one is GCD
*/
@@ -355,9 +364,22 @@
slave_port_id);
if (find_slave_by_id(internals->active_slaves,
- internals->active_slave_count,
- slave_port_id) == internals->active_slave_count)
+ internals->active_slave_count,
+ slave_port_id) == internals->active_slave_count) {
+ if (internals->mode4.slave_link_valid == 0) {
+ internals->mode4.slave_link_valid = 1;
+ /* Inherit dev link properties from first active slave */
+ link_properties_set(bonded_eth_dev,
+ &(slave_eth_dev->data->dev_link));
+ } else if (link_properties_valid(bonded_eth_dev,
+ &slave_eth_dev->data->dev_link) != 0) {
+ RTE_BOND_LOG(ERR, "Invalid link properties for slave %d"
+ " in bonding mode %d", slave_port_id,
+ internals->mode);
+ return -1;
+ }
activate_slave(bonded_eth_dev, slave_port_id);
+ }
}
}
--
1.9.1
next reply other threads:[~2017-11-29 14:55 UTC|newest]
Thread overview: 8+ messages / expand[flat|nested] mbox.gz Atom feed top
2017-11-29 14:53 Tomasz Kulasek [this message]
2017-11-29 15:42 ` [dpdk-dev] [PATCH v2] " Tomasz Kulasek
2018-01-17 16:02 ` Ferruh Yigit
2018-02-06 20:52 ` Thomas Monjalon
2018-02-12 18:26 ` Chas Williams
2018-06-13 16:10 ` Ferruh Yigit
2018-06-18 8:39 ` Ferruh Yigit
2018-02-16 20:13 ` Stephen Hemminger
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=20171129145349.43448-1-tomaszx.kulasek@intel.com \
--to=tomaszx.kulasek@intel.com \
--cc=declan.doherty@intel.com \
--cc=dev@dpdk.org \
/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).