lists.openwall.net   lists  /  announce  owl-users  owl-dev  john-users  john-dev  passwdqc-users  yescrypt  popa3d-users  /  oss-security  kernel-hardening  musl  sabotage  tlsify  passwords  /  crypt-dev  xvendor  /  Bugtraq  Full-Disclosure  linux-kernel  linux-netdev  linux-ext4  linux-hardening  PHC 
Open Source and information security mailing list archives
 
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date:   Thu, 3 Sep 2020 22:34:57 +0100
From:   Edward Cree <ecree@...arflare.com>
To:     <linux-net-drivers@...arflare.com>, <davem@...emloft.net>
CC:     <netdev@...r.kernel.org>
Subject: [PATCH v2 net-next 3/6] sfc: use tx_queue->old_read_count in EF100 TX
 path

As in the Siena/EF10 case, it minimises cacheline ping-pong between
 the TX and completion paths.

Signed-off-by: Edward Cree <ecree@...arflare.com>
---
 drivers/net/ethernet/sfc/ef100_tx.c   |  8 ++++++--
 drivers/net/ethernet/sfc/net_driver.h | 14 ++++++++++++++
 2 files changed, 20 insertions(+), 2 deletions(-)

diff --git a/drivers/net/ethernet/sfc/ef100_tx.c b/drivers/net/ethernet/sfc/ef100_tx.c
index ce1b462efd17..078c7ec2a70e 100644
--- a/drivers/net/ethernet/sfc/ef100_tx.c
+++ b/drivers/net/ethernet/sfc/ef100_tx.c
@@ -360,15 +360,19 @@ int ef100_enqueue_skb(struct efx_tx_queue *tx_queue, struct sk_buff *skb)
 		goto err;
 	ef100_tx_make_descriptors(tx_queue, skb, segments);
 
-	fill_level = efx_channel_tx_fill_level(tx_queue->channel);
+	fill_level = efx_channel_tx_old_fill_level(tx_queue->channel);
 	if (fill_level > efx->txq_stop_thresh) {
+		struct efx_tx_queue *txq2;
+
 		netif_tx_stop_queue(tx_queue->core_txq);
 		/* Re-read after a memory barrier in case we've raced with
 		 * the completion path. Otherwise there's a danger we'll never
 		 * restart the queue if all completions have just happened.
 		 */
 		smp_mb();
-		fill_level = efx_channel_tx_fill_level(tx_queue->channel);
+		efx_for_each_channel_tx_queue(txq2, tx_queue->channel)
+			txq2->old_read_count = READ_ONCE(txq2->read_count);
+		fill_level = efx_channel_tx_old_fill_level(tx_queue->channel);
 		if (fill_level < efx->txq_stop_thresh)
 			netif_tx_start_queue(tx_queue->core_txq);
 	}
diff --git a/drivers/net/ethernet/sfc/net_driver.h b/drivers/net/ethernet/sfc/net_driver.h
index adc138f9d15f..366e649fa869 100644
--- a/drivers/net/ethernet/sfc/net_driver.h
+++ b/drivers/net/ethernet/sfc/net_driver.h
@@ -1692,6 +1692,20 @@ efx_channel_tx_fill_level(struct efx_channel *channel)
 	return fill_level;
 }
 
+/* Conservative approximation of efx_channel_tx_fill_level using cached value */
+static inline unsigned int
+efx_channel_tx_old_fill_level(struct efx_channel *channel)
+{
+	struct efx_tx_queue *tx_queue;
+	unsigned int fill_level = 0;
+
+	efx_for_each_channel_tx_queue(tx_queue, channel)
+		fill_level = max(fill_level,
+				 tx_queue->insert_count - tx_queue->old_read_count);
+
+	return fill_level;
+}
+
 /* Get all supported features.
  * If a feature is not fixed, it is present in hw_features.
  * If a feature is fixed, it does not present in hw_features, but

Powered by blists - more mailing lists