The ice_aq_get_link_info function is not thread-safe. However,
it is possible to simultaneous invocations during both the dev_start
and the LSC interrupt handler, potentially leading to unexpected adminq
errors. This patch addresses the issue by introducing a thread-safe
wrapper that utilizes a spinlock.
Fixes: cf911d90e366 ("net/ice: support link update")
Cc: sta...@dpdk.org
Signed-off-by: Qi Zhang <qi.z.zh...@intel.com>
---
drivers/net/ice/ice_ethdev.c | 26 ++++++++++++++++++++------
drivers/net/ice/ice_ethdev.h | 3 +++
2 files changed, 23 insertions(+), 6 deletions(-)
diff --git a/drivers/net/ice/ice_ethdev.c b/drivers/net/ice/ice_ethdev.c
index 3ccba4db80..1f8ab5158a 100644
--- a/drivers/net/ice/ice_ethdev.c
+++ b/drivers/net/ice/ice_ethdev.c
@@ -1804,6 +1804,7 @@ ice_pf_setup(struct ice_pf *pf)
}
pf->main_vsi = vsi;
+ rte_spinlock_init(&pf->link_lock);
return 0;
}
@@ -3621,17 +3622,31 @@ ice_rxq_intr_setup(struct rte_eth_dev *dev)
return 0;
}
+static enum ice_status
+ice_get_link_info_safe(struct ice_pf *pf, bool ena_lse,
+ struct ice_link_status *link)
+{
+ struct ice_hw *hw = ICE_PF_TO_HW(pf);
+ int ret;
+
+ rte_spinlock_lock(&pf->link_lock);
+
+ ret = ice_aq_get_link_info(hw->port_info, ena_lse, link, NULL);
+
+ rte_spinlock_unlock(&pf->link_lock);
+
+ return ret;
+}
+
static void
ice_get_init_link_status(struct rte_eth_dev *dev)
{
- struct ice_hw *hw = ICE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
struct ice_pf *pf = ICE_DEV_PRIVATE_TO_PF(dev->data->dev_private);
bool enable_lse = dev->data->dev_conf.intr_conf.lsc ? true : false;
struct ice_link_status link_status;
int ret;
- ret = ice_aq_get_link_info(hw->port_info, enable_lse,
- &link_status, NULL);
+ ret = ice_get_link_info_safe(pf, enable_lse, &link_status);
if (ret != ICE_SUCCESS) {
PMD_DRV_LOG(ERR, "Failed to get link info");
pf->init_link_up = false;
@@ -3996,7 +4011,7 @@ ice_link_update(struct rte_eth_dev *dev, int
wait_to_complete)
{
#define CHECK_INTERVAL 50 /* 50ms */
#define MAX_REPEAT_TIME 40 /* 2s (40 * 50ms) in total */
- struct ice_hw *hw = ICE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+ struct ice_pf *pf = ICE_DEV_PRIVATE_TO_PF(dev->data->dev_private);
struct ice_link_status link_status;
struct rte_eth_link link, old;
int status;
@@ -4010,8 +4025,7 @@ ice_link_update(struct rte_eth_dev *dev, int
wait_to_complete)
do {
/* Get link status information from hardware */
- status = ice_aq_get_link_info(hw->port_info, enable_lse,
- &link_status, NULL);
+ status = ice_get_link_info_safe(pf, enable_lse, &link_status);
if (status != ICE_SUCCESS) {
link.link_speed = RTE_ETH_SPEED_NUM_100M;
link.link_duplex = RTE_ETH_LINK_FULL_DUPLEX;
diff --git a/drivers/net/ice/ice_ethdev.h b/drivers/net/ice/ice_ethdev.h
index abe6dcdc23..691893be13 100644
--- a/drivers/net/ice/ice_ethdev.h
+++ b/drivers/net/ice/ice_ethdev.h
@@ -548,6 +548,9 @@ struct ice_pf {
uint64_t rss_hf;
struct ice_tm_conf tm_conf;
uint16_t outer_ethertype;
+ /* lock prevent race condition between lsc interrupt handler
+ * and link status update during dev_start */
+ rte_spinlock_t link_lock;
};
#define ICE_MAX_QUEUE_NUM 2048
--
2.31.1
