Skip to contentSkip to conversation

bnx2: Fix lost MSI-X problem on 5709 NICs.

PATCH Applied Michael Chan 2 participants 6 messages Apr 27, 2010

Subsystems (2)
Networking Network drivers
lore
Share & tools
lore

Michael Chan

author
#0 lore raw

PATCH 1/3 bnx2: Fix lost MSI-X problem on 5709 NICs.

It has been reported that under certain heavy traffic conditions in MSI-X mode, the driver can lose an MSI-X vector causing all packets in the associated rx/tx ring pair to be dropped. The problem is caused by the chip dropping the write to unmask the MSI-X vector by the kernel (when migrating the IRQ for example).

This can be prevented by increasing the GRC timeout value for these register read and write operations.

Thanks to Dell for helping us debug this problem.

Signed-off-by Michael Chan <[email protected]>
---
 drivers/net/bnx2.c |    6 +++++-
 1 files changed, 5 insertions(+), 1 deletions(-)
drivers/net/bnx2.c +5 -1
index a257bab..4c1e51e 100644
@@ -4759,8 +4759,12 @@ bnx2_reset_chip(struct bnx2 *bp, u32 reset_code)
 		rc = bnx2_alloc_bad_rbuf(bp);
 	}
 
-	if (bp->flags & BNX2_FLAG_USING_MSIX)
+	if (bp->flags & BNX2_FLAG_USING_MSIX) {
 		bnx2_setup_msix_tbl(bp);
+		/* Prevent MSIX table reads and write from timing out */
+		REG_WR(bp, BNX2_MISC_ECO_HW_CTL,
+			BNX2_MISC_ECO_HW_CTL_LARGE_GRC_TMOUT_EN);
+	}
 
 	return rc;
 }
Signature · 1 line
1.6.4.GIT
author
#1 lore raw

PATCH 2/3 bnx2: Prevent "scheduling while atomic" warning with cnic, bonding and vlan.

The bonding driver calls ndo_vlan_rx_register() while holding bond->lock. The bnx2 driver calls bnx2_netif_stop() to stop the rx handling while changing the vlgrp. The call also stops the cnic driver which sleeps while the bond->lock is held and cause the warning.

This code path only needs to stop the NAPI rx handling while we are changing the vlgrp. Since no reset is going to occur, there is no need to stop cnic in this case. By adding a parameter to bnx2_netif_stop() to skip stopping cnic, we can avoid the warning.

Signed-off-by Michael Chan <[email protected]>
---
 drivers/net/bnx2.c |   38 ++++++++++++++++++++------------------
 1 files changed, 20 insertions(+), 18 deletions(-)
drivers/net/bnx2.c +20 -18
index 4c1e51e..35eec2d 100644
@@ -651,9 +651,10 @@ bnx2_napi_enable(struct bnx2 *bp)
 }
 
 static void
-bnx2_netif_stop(struct bnx2 *bp)
+bnx2_netif_stop(struct bnx2 *bp, bool stop_cnic)
 {
-	bnx2_cnic_stop(bp);
+	if (stop_cnic)
+		bnx2_cnic_stop(bp);
 	if (netif_running(bp->dev)) {
 		int i;
 

@@ -671,14 +672,15 @@ bnx2_netif_stop(struct bnx2 *bp)
 }
 
 static void
-bnx2_netif_start(struct bnx2 *bp)
+bnx2_netif_start(struct bnx2 *bp, bool start_cnic)
 {
 	if (atomic_dec_and_test(&bp->intr_sem)) {
 		if (netif_running(bp->dev)) {
 			netif_tx_wake_all_queues(bp->dev);
 			bnx2_napi_enable(bp);
 			bnx2_enable_int(bp);
-			bnx2_cnic_start(bp);
+			if (start_cnic)
+				bnx2_cnic_start(bp);
 		}
 	}
 }

@@ -6277,12 +6279,12 @@ bnx2_reset_task(struct work_struct *work)
 		return;
 	}
 
-	bnx2_netif_stop(bp);
+	bnx2_netif_stop(bp, true);
 
 	bnx2_init_nic(bp, 1);
 
 	atomic_set(&bp->intr_sem, 1);
-	bnx2_netif_start(bp);
+	bnx2_netif_start(bp, true);
 	rtnl_unlock();
 }
 

@@ -6324,7 +6326,7 @@ bnx2_vlan_rx_register(struct net_device *dev, struct vlan_group *vlgrp)
 	struct bnx2 *bp = netdev_priv(dev);
 
 	if (netif_running(dev))
-		bnx2_netif_stop(bp);
+		bnx2_netif_stop(bp, false);
 
 	bp->vlgrp = vlgrp;
 

@@ -6335,7 +6337,7 @@ bnx2_vlan_rx_register(struct net_device *dev, struct vlan_group *vlgrp)
 	if (bp->flags & BNX2_FLAG_CAN_KEEP_VLAN)
 		bnx2_fw_sync(bp, BNX2_DRV_MSG_CODE_KEEP_VLAN_UPDATE, 0, 1);
 
-	bnx2_netif_start(bp);
+	bnx2_netif_start(bp, false);
 }
 #endif
 

@@ -7055,9 +7057,9 @@ bnx2_set_coalesce(struct net_device *dev, struct ethtool_coalesce *coal)
 	bp->stats_ticks &= BNX2_HC_STATS_TICKS_HC_STAT_TICKS;
 
 	if (netif_running(bp->dev)) {
-		bnx2_netif_stop(bp);
+		bnx2_netif_stop(bp, true);
 		bnx2_init_nic(bp, 0);
-		bnx2_netif_start(bp);
+		bnx2_netif_start(bp, true);
 	}
 
 	return 0;

@@ -7087,7 +7089,7 @@ bnx2_change_ring_size(struct bnx2 *bp, u32 rx, u32 tx)
 		/* Reset will erase chipset stats; save them */
 		bnx2_save_stats(bp);
 
-		bnx2_netif_stop(bp);
+		bnx2_netif_stop(bp, true);
 		bnx2_reset_chip(bp, BNX2_DRV_MSG_CODE_RESET);
 		bnx2_free_skbs(bp);
 		bnx2_free_mem(bp);

@@ -7115,7 +7117,7 @@ bnx2_change_ring_size(struct bnx2 *bp, u32 rx, u32 tx)
 			bnx2_setup_cnic_irq_info(bp);
 		mutex_unlock(&bp->cnic_lock);
 #endif
-		bnx2_netif_start(bp);
+		bnx2_netif_start(bp, true);
 	}
 	return 0;
 }

@@ -7368,7 +7370,7 @@ bnx2_self_test(struct net_device *dev, struct ethtool_test *etest, u64 *buf)
 	if (etest->flags & ETH_TEST_FL_OFFLINE) {
 		int i;
 
-		bnx2_netif_stop(bp);
+		bnx2_netif_stop(bp, true);
 		bnx2_reset_chip(bp, BNX2_DRV_MSG_CODE_DIAG);
 		bnx2_free_skbs(bp);
 

@@ -7387,7 +7389,7 @@ bnx2_self_test(struct net_device *dev, struct ethtool_test *etest, u64 *buf)
 			bnx2_shutdown_chip(bp);
 		else {
 			bnx2_init_nic(bp, 1);
-			bnx2_netif_start(bp);
+			bnx2_netif_start(bp, true);
 		}
 
 		/* wait for link up */

@@ -8381,7 +8383,7 @@ bnx2_suspend(struct pci_dev *pdev, pm_message_t state)
 		return 0;
 
 	flush_scheduled_work();
-	bnx2_netif_stop(bp);
+	bnx2_netif_stop(bp, true);
 	netif_device_detach(dev);
 	del_timer_sync(&bp->timer);
 	bnx2_shutdown_chip(bp);

@@ -8403,7 +8405,7 @@ bnx2_resume(struct pci_dev *pdev)
 	bnx2_set_power_state(bp, PCI_D0);
 	netif_device_attach(dev);
 	bnx2_init_nic(bp, 1);
-	bnx2_netif_start(bp);
+	bnx2_netif_start(bp, true);
 	return 0;
 }
 

@@ -8430,7 +8432,7 @@ static pci_ers_result_t bnx2_io_error_detected(struct pci_dev *pdev,
 	}
 
 	if (netif_running(dev)) {
-		bnx2_netif_stop(bp);
+		bnx2_netif_stop(bp, true);
 		del_timer_sync(&bp->timer);
 		bnx2_reset_nic(bp, BNX2_DRV_MSG_CODE_RESET);
 	}

@@ -8487,7 +8489,7 @@ static void bnx2_io_resume(struct pci_dev *pdev)
 
 	rtnl_lock();
 	if (netif_running(dev))
-		bnx2_netif_start(bp);
+		bnx2_netif_start(bp, true);
 
 	netif_device_attach(dev);
 	rtnl_unlock();
Signature · 1 line
1.6.4.GIT
author
#2 lore raw

PATCH 3/3 bnx2: Update version to 2.0.9.

Signed-off-by Michael Chan <[email protected]>
---
 drivers/net/bnx2.c |    4 ++--
 1 files changed, 2 insertions(+), 2 deletions(-)
drivers/net/bnx2.c +2 -2
index 35eec2d..ac90a38 100644
@@ -58,8 +58,8 @@
 #include "bnx2_fw.h"
 
 #define DRV_MODULE_NAME		"bnx2"
-#define DRV_MODULE_VERSION	"2.0.8"
-#define DRV_MODULE_RELDATE	"Feb 15, 2010"
+#define DRV_MODULE_VERSION	"2.0.9"
+#define DRV_MODULE_RELDATE	"April 27, 2010"
 #define FW_MIPS_FILE_06		"bnx2/bnx2-mips-06-5.0.0.j6.fw"
 #define FW_RV2P_FILE_06		"bnx2/bnx2-rv2p-06-5.0.0.j3.fw"
 #define FW_MIPS_FILE_09		"bnx2/bnx2-mips-09-5.0.0.j9.fw"
Signature · 1 line
1.6.4.GIT
maintainer Applied
#4 lore raw

Re: [PATCH 2/3] bnx2: Prevent "scheduling while atomic" warning with cnic, bonding and vlan.

In reply to Michael Chan

From: "Michael Chan" <[email protected]>
Date: Tue, 27 Apr 2010 14:28:10 -0700

Quoted 11 lines

The bonding driver calls ndo_vlan_rx_register() while holding bond->lock. The bnx2 driver calls bnx2_netif_stop() to stop the rx handling while changing the vlgrp. The call also stops the cnic driver which sleeps while the bond->lock is held and cause the warning.

This code path only needs to stop the NAPI rx handling while we are changing the vlgrp. Since no reset is going to occur, there is no need to stop cnic in this case. By adding a parameter to bnx2_netif_stop() to skip stopping cnic, we can avoid the warning.

Signed-off-by Michael Chan <[email protected]>

Applied to net-2.6

maintainer Applied
#5 lore raw
In reply to Michael Chan

From: "Michael Chan" <[email protected]>
Date: Tue, 27 Apr 2010 14:28:09 -0700

Quoted 6 lines

It has been reported that under certain heavy traffic conditions in MSI-X mode, the driver can lose an MSI-X vector causing all packets in the associated rx/tx ring pair to be dropped. The problem is caused by the chip dropping the write to unmask the MSI-X vector by the kernel (when migrating the IRQ for example).

Quoted 6 lines

This can be prevented by increasing the GRC timeout value for these register read and write operations.

Thanks to Dell for helping us debug this problem.

Signed-off-by Michael Chan <[email protected]>

Applied to net-2.6