After rte_eth_dev_set_link_down() the port can come back up without a
corresponding rte_eth_dev_set_link_up(). Three paths reprogram the PHY
while the port is down:
- txgbe_dev_e56_check_bp_event() (the AN73 watchdog): keeps driving
negotiation while the link is down and auto-negotiation is enabled.
- txgbe_dev_setup_link_alarm_handler_aml(): runs setup_link() from the
2 s retry armed by txgbe_setup_mac_link_aml() and from the link
thread.
- txgbe_dev_detect_sfp(): calls txgbe_dev_setup_link_alarm_handler(),
which ignores the administrative state, after identifying a module.
The first two are gated on hw->dev_start before acquire_swfw_sync() and
setup_link(); the third only around the setup call, so identification
keeps running. It has to keep running: no setup path re-identifies the
module by itself, since txgbe_setup_phy_link_aml() never identifies and
txgbe_setup_phy_link_aml40() identifies only when the cached type is
not present. A module swapped while the port is down would otherwise be
configured as the module that was there before, out of hw->phy.sfp_type.
set_link_down() then stops what is already running:
- clears hw->dev_start first, so a callback that starts later stops at
its gate;
- calls rte_eal_alarm_cancel(), which returns only after a matching
callback that is already running has finished;
- waits for the link thread with txgbe_dev_wait_setup_link_complete(),
which with timeout 0 has no bound and logs an error every 9 s;
- cancels the handler once more: the link thread runs the same handler
and can arm a 2 s retry that the earlier cancel never saw;
- calls hw->mac.disable_tx_laser() last; on a 40G DAC that clears the
transmit and receive enables in PMD_CFG0.
set_link_up() restarts the flow for both Amber-Lite MACs. The watchdog
is cancelled before it is armed because rte_eal_alarm_set() adds an
entry on every call without looking for an existing one, and each entry
arms the next when it fires: arming twice leaves two chains running
side by side. Link setup now runs for txgbe_mac_aml too and returns
-EIO on failure; txgbe_setup_phy_link_aml() resets the PHY and restarts
AN73 (when auto-negotiation is enabled).
Fixes: 234ce0d1fa9d ("net/txgbe: fix link stability for Amber-Lite backplane
mode")
Fixes: 6104fd11086e ("net/txgbe: fix link stability for 25G NIC")
Cc: [email protected]
Signed-off-by: Zaiyu Wang <[email protected]>
---
drivers/net/txgbe/txgbe_ethdev.c | 40 ++++++++++++++++++++++++++++----
1 file changed, 35 insertions(+), 5 deletions(-)
diff --git a/drivers/net/txgbe/txgbe_ethdev.c b/drivers/net/txgbe/txgbe_ethdev.c
index 043c96b747..9dbdbfa309 100644
--- a/drivers/net/txgbe/txgbe_ethdev.c
+++ b/drivers/net/txgbe/txgbe_ethdev.c
@@ -2213,7 +2213,8 @@ txgbe_dev_set_link_up(struct rte_eth_dev *dev)
/* Turn on the laser */
hw->mac.enable_tx_laser(hw);
- if (hw->mac.type == txgbe_mac_aml40) {
+ if (hw->mac.type == txgbe_mac_aml ||
+ hw->mac.type == txgbe_mac_aml40) {
speed = hw->phy.autoneg_advertised;
if (speed == 0) {
err = hw->mac.get_link_capabilities(hw, &speed,
&autoneg);
@@ -2227,12 +2228,33 @@ txgbe_dev_set_link_up(struct rte_eth_dev *dev)
}
hw->dev_start = true;
+ if (hw->mac.type == txgbe_mac_aml ||
+ hw->mac.type == txgbe_mac_aml40) {
+ rte_eal_alarm_cancel(txgbe_dev_e56_check_bp_event, dev);
+ rte_eal_alarm_set(hw->bp_event_interval,
+ txgbe_dev_e56_check_bp_event, dev);
+ }
txgbe_dev_link_update(dev, 0);
}
return 0;
}
+static void
+txgbe_stop_link_handlers(struct rte_eth_dev *dev)
+{
+ struct txgbe_hw *hw = TXGBE_DEV_HW(dev);
+
+ if (hw->mac.type != txgbe_mac_aml &&
+ hw->mac.type != txgbe_mac_aml40)
+ return;
+
+ rte_eal_alarm_cancel(txgbe_dev_e56_check_bp_event, dev);
+ rte_eal_alarm_cancel(txgbe_dev_setup_link_alarm_handler_aml, hw);
+ txgbe_dev_wait_setup_link_complete(dev, 0);
+ rte_eal_alarm_cancel(txgbe_dev_setup_link_alarm_handler_aml, hw);
+}
+
/*
* Set device link down: disable tx.
*/
@@ -2245,9 +2267,11 @@ txgbe_dev_set_link_down(struct rte_eth_dev *dev)
/* Turn off the copper */
hw->phy.set_phy_power(hw, false);
} else {
+ hw->dev_start = false;
+ txgbe_stop_link_handlers(dev);
+
/* Turn off the laser */
hw->mac.disable_tx_laser(hw);
- hw->dev_start = false;
txgbe_dev_link_update(dev, 0);
}
@@ -3073,8 +3097,8 @@ void txgbe_dev_e56_check_bp_event(void *param)
if (!hw->devarg.auto_neg)
return;
- /* only continue if link is down */
- if (dev->data->dev_link.link_status)
+ /* Continue only for an administratively up port with link down. */
+ if (!hw->dev_start || dev->data->dev_link.link_status)
goto out;
value = rd32_epcs(hw, VR_AN_INTR);
@@ -3232,7 +3256,8 @@ txgbe_dev_detect_sfp(void *param)
} else if (err == 0) {
hw->mac.setup_sfp(hw);
PMD_DRV_LOG(INFO, "detected SFP+: %d", hw->phy.sfp_type);
- txgbe_dev_setup_link_alarm_handler(dev);
+ if (hw->dev_start)
+ txgbe_dev_setup_link_alarm_handler(dev);
/* Re-arm the AN73 watchdog for the newly inserted module, so
* that only one instance of it is running at a time.
*/
@@ -3412,6 +3437,11 @@ txgbe_dev_setup_link_alarm_handler_aml(void *param)
bool autoneg = false;
u32 gssr = hw->phy.phy_semaphore_mask;
+ if (!hw->dev_start) {
+ intr->flags &= ~TXGBE_FLAG_NEED_LINK_CONFIG;
+ return;
+ }
+
speed = hw->phy.autoneg_advertised;
if (!speed)
hw->mac.get_link_capabilities(hw, &speed, &autoneg);
--
2.55.0.windows.2
