diff mbox series

[V2,net-next] liquidio: prevent rx queues from getting stalled

Message ID 20180329181322.GA9939@felix-thinkpad.cavium.com
State Accepted, archived
Delegated to: David Miller
Headers show
Series [V2,net-next] liquidio: prevent rx queues from getting stalled | expand

Commit Message

Manlunas, Felix March 29, 2018, 6:13 p.m. UTC 
From: Raghu Vatsavayi <raghu.vatsavayi@cavium.com>

This commit has fix for RX traffic issues when we stress test the driver
with continuous ifconfig up/down under very high traffic conditions.

Reason for the issue is that, in existing liquidio_stop function NAPI is
disabled even before actual FW/HW interface is brought down via
send_rx_ctrl_cmd(lio, 0). Between time frame of NAPI disable and actual
interface down in firmware, firmware continuously enqueues rx traffic to
host. When interrupt happens for new packets, host irq handler fails in
scheduling NAPI as the NAPI is already disabled.

After "ifconfig <iface> up", Host re-enables NAPI but cannot schedule it
until it receives another Rx interrupt. Host never receives Rx interrupt as
it never cleared the Rx interrupt it received during interface down
operation. NIC Rx interrupt gets cleared only when Host processes queue and
clears the queue counts. Above anomaly leads to other issues like packet
overflow in FW/HW queues, backpressure.

Fix:
This commit fixes this issue by disabling NAPI only after informing
firmware to stop queueing packets to host via send_rx_ctrl_cmd(lio, 0).
send_rx_ctrl_cmd is not visible in the patch as it is already there in the
code. The DOWN command also waits for any pending packets to be processed
by NAPI so that the deadlock will not occur.

Signed-off-by: Raghu Vatsavayi <raghu.vatsavayi@cavium.com>
Acked-by: Derek Chickles <derek.chickles@cavium.com>
Signed-off-by: Felix Manlunas <felix.manlunas@cavium.com>
---
Patch Change Log:
  V1 -> V2:
    Revised the subject line and patch description to add clarity.
    No change in the actual code.

 drivers/net/ethernet/cavium/liquidio/lio_core.c    | 23 ++++++++++++++++++++
 drivers/net/ethernet/cavium/liquidio/lio_main.c    | 25 +++++++++++++---------
 drivers/net/ethernet/cavium/liquidio/lio_vf_main.c | 23 ++++++++++++--------
 .../net/ethernet/cavium/liquidio/octeon_network.h  |  1 +
 4 files changed, 53 insertions(+), 19 deletions(-)

Comments

David Miller March 30, 2018, 6:21 p.m. UTC | #1 
From: Felix Manlunas <felix.manlunas@cavium.com>
Date: Thu, 29 Mar 2018 11:13:22 -0700

> From: Raghu Vatsavayi <raghu.vatsavayi@cavium.com>
> 
> This commit has fix for RX traffic issues when we stress test the driver
> with continuous ifconfig up/down under very high traffic conditions.
> 
> Reason for the issue is that, in existing liquidio_stop function NAPI is
> disabled even before actual FW/HW interface is brought down via
> send_rx_ctrl_cmd(lio, 0). Between time frame of NAPI disable and actual
> interface down in firmware, firmware continuously enqueues rx traffic to
> host. When interrupt happens for new packets, host irq handler fails in
> scheduling NAPI as the NAPI is already disabled.
> 
> After "ifconfig <iface> up", Host re-enables NAPI but cannot schedule it
> until it receives another Rx interrupt. Host never receives Rx interrupt as
> it never cleared the Rx interrupt it received during interface down
> operation. NIC Rx interrupt gets cleared only when Host processes queue and
> clears the queue counts. Above anomaly leads to other issues like packet
> overflow in FW/HW queues, backpressure.
> 
> Fix:
> This commit fixes this issue by disabling NAPI only after informing
> firmware to stop queueing packets to host via send_rx_ctrl_cmd(lio, 0).
> send_rx_ctrl_cmd is not visible in the patch as it is already there in the
> code. The DOWN command also waits for any pending packets to be processed
> by NAPI so that the deadlock will not occur.
> 
> Signed-off-by: Raghu Vatsavayi <raghu.vatsavayi@cavium.com>
> Acked-by: Derek Chickles <derek.chickles@cavium.com>
> Signed-off-by: Felix Manlunas <felix.manlunas@cavium.com>

Applied.
diff mbox series

Patch

diff --git a/drivers/net/ethernet/cavium/liquidio/lio_core.c b/drivers/net/ethernet/cavium/liquidio/lio_core.c
index 73e70e0..2a94eee 100644
--- a/drivers/net/ethernet/cavium/liquidio/lio_core.c
+++ b/drivers/net/ethernet/cavium/liquidio/lio_core.c
@@ -1146,3 +1146,26 @@  int liquidio_change_mtu(struct net_device *netdev, int new_mtu)
 	octeon_free_soft_command(oct, sc);
 	return 0;
 }
+
+int lio_wait_for_clean_oq(struct octeon_device *oct)
+{
+	int retry = 100, pending_pkts = 0;
+	int idx;
+
+	do {
+		pending_pkts = 0;
+
+		for (idx = 0; idx < MAX_OCTEON_OUTPUT_QUEUES(oct); idx++) {
+			if (!(oct->io_qmask.oq & BIT_ULL(idx)))
+				continue;
+			pending_pkts +=
+				atomic_read(&oct->droq[idx]->pkts_pending);
+		}
+
+		if (pending_pkts > 0)
+			schedule_timeout_uninterruptible(1);
+
+	} while (retry-- && pending_pkts);
+
+	return pending_pkts;
+}
diff --git a/drivers/net/ethernet/cavium/liquidio/lio_main.c b/drivers/net/ethernet/cavium/liquidio/lio_main.c
index 43c5ba0..603a144 100644
--- a/drivers/net/ethernet/cavium/liquidio/lio_main.c
+++ b/drivers/net/ethernet/cavium/liquidio/lio_main.c
@@ -2084,16 +2084,6 @@  static int liquidio_stop(struct net_device *netdev)
 	struct octeon_device *oct = lio->oct_dev;
 	struct napi_struct *napi, *n;
 
-	if (oct->props[lio->ifidx].napi_enabled) {
-		list_for_each_entry_safe(napi, n, &netdev->napi_list, dev_list)
-			napi_disable(napi);
-
-		oct->props[lio->ifidx].napi_enabled = 0;
-
-		if (OCTEON_CN23XX_PF(oct))
-			oct->droq[0]->ops.poll_mode = 0;
-	}
-
 	ifstate_reset(lio, LIO_IFSTATE_RUNNING);
 
 	netif_tx_disable(netdev);
@@ -2119,6 +2109,21 @@  static int liquidio_stop(struct net_device *netdev)
 		lio->ptp_clock = NULL;
 	}
 
+	/* Wait for any pending Rx descriptors */
+	if (lio_wait_for_clean_oq(oct))
+		netif_info(lio, rx_err, lio->netdev,
+			   "Proceeding with stop interface after partial RX desc processing\n");
+
+	if (oct->props[lio->ifidx].napi_enabled == 1) {
+		list_for_each_entry_safe(napi, n, &netdev->napi_list, dev_list)
+			napi_disable(napi);
+
+		oct->props[lio->ifidx].napi_enabled = 0;
+
+		if (OCTEON_CN23XX_PF(oct))
+			oct->droq[0]->ops.poll_mode = 0;
+	}
+
 	dev_info(&oct->pci_dev->dev, "%s interface is stopped\n", netdev->name);
 
 	return 0;
diff --git a/drivers/net/ethernet/cavium/liquidio/lio_vf_main.c b/drivers/net/ethernet/cavium/liquidio/lio_vf_main.c
index dc62698..f92dfa4 100644
--- a/drivers/net/ethernet/cavium/liquidio/lio_vf_main.c
+++ b/drivers/net/ethernet/cavium/liquidio/lio_vf_main.c
@@ -1138,15 +1138,6 @@  static int liquidio_stop(struct net_device *netdev)
 	/* tell Octeon to stop forwarding packets to host */
 	send_rx_ctrl_cmd(lio, 0);
 
-	if (oct->props[lio->ifidx].napi_enabled) {
-		list_for_each_entry_safe(napi, n, &netdev->napi_list, dev_list)
-			napi_disable(napi);
-
-		oct->props[lio->ifidx].napi_enabled = 0;
-
-		oct->droq[0]->ops.poll_mode = 0;
-	}
-
 	netif_info(lio, ifdown, lio->netdev, "Stopping interface!\n");
 	/* Inform that netif carrier is down */
 	lio->intf_open = 0;
@@ -1159,6 +1150,20 @@  static int liquidio_stop(struct net_device *netdev)
 
 	stop_txqs(netdev);
 
+	/* Wait for any pending Rx descriptors */
+	if (lio_wait_for_clean_oq(oct))
+		netif_info(lio, rx_err, lio->netdev,
+			   "Proceeding with stop interface after partial RX desc processing\n");
+
+	if (oct->props[lio->ifidx].napi_enabled == 1) {
+		list_for_each_entry_safe(napi, n, &netdev->napi_list, dev_list)
+			napi_disable(napi);
+
+		oct->props[lio->ifidx].napi_enabled = 0;
+
+		oct->droq[0]->ops.poll_mode = 0;
+	}
+
 	dev_info(&oct->pci_dev->dev, "%s interface is stopped\n", netdev->name);
 
 	return 0;
diff --git a/drivers/net/ethernet/cavium/liquidio/octeon_network.h b/drivers/net/ethernet/cavium/liquidio/octeon_network.h
index 8782206..4069710 100644
--- a/drivers/net/ethernet/cavium/liquidio/octeon_network.h
+++ b/drivers/net/ethernet/cavium/liquidio/octeon_network.h
@@ -190,6 +190,7 @@  irqreturn_t liquidio_msix_intr_handler(int irq __attribute__((unused)),
 
 int octeon_setup_interrupt(struct octeon_device *oct, u32 num_ioqs);
 
+int lio_wait_for_clean_oq(struct octeon_device *oct);
 /**
  * \brief Register ethtool operations
  * @param netdev    pointer to network device