From: Yuanhan Liu <yuanhan.liu@linux.intel.com>
To: Kevin Traynor <ktraynor@redhat.com>
Cc: Yuanhan Liu <yuanhan.liu@linux.intel.com>,
Maxime Coquelin <maxime.coquelin@redhat.com>,
dpdk stable <stable@dpdk.org>
Subject: [dpdk-stable] patch 'vhost: fix false sharing' has been queued to LTS release 16.11.2
Date: Fri, 7 Apr 2017 16:11:56 +0800 [thread overview]
Message-ID: <1491552724-3034-39-git-send-email-yuanhan.liu@linux.intel.com> (raw)
In-Reply-To: <1491552724-3034-1-git-send-email-yuanhan.liu@linux.intel.com>
Hi,
FYI, your patch has been queued to LTS release 16.11.2
Note it hasn't been pushed to http://dpdk.org/browse/dpdk-stable
yet. It will be pushed if I get no objections before 04/11/17.
So please shout if anyone has objections.
Thanks.
--yliu
---
>From 071a0f1441a087edc4337bdc2ea9f4e038099973 Mon Sep 17 00:00:00 2001
From: Kevin Traynor <ktraynor@redhat.com>
Date: Thu, 23 Mar 2017 15:44:58 +0000
Subject: [PATCH] vhost: fix false sharing
[ upstream commit 4e9474141e1115a3089896848f92b25b53cddfd9 ]
The broadcast_rarp field in the virtio_net struct is checked in the
dequeue datapath regardless of whether descriptors are available or not.
As it is checked with cmpset leading to a write, false sharing on the
virtio_net struct can happen between enqueue and dequeue datapaths
regardless of whether a RARP is requested. In OVS, the issue can cause
a uni-directional performance drop of up to 15%.
Fix that by only performing the cmpset if a read of broadcast_rarp
indicates that the cmpset is likely to succeed.
Fixes: a66bcad32240 ("vhost: arrange struct fields for better cache sharing")
Signed-off-by: Kevin Traynor <ktraynor@redhat.com>
Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
Acked-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
---
lib/librte_vhost/virtio_net.c | 16 ++++++++++++++--
1 file changed, 14 insertions(+), 2 deletions(-)
diff --git a/lib/librte_vhost/virtio_net.c b/lib/librte_vhost/virtio_net.c
index 337470d..d0a3b11 100644
--- a/lib/librte_vhost/virtio_net.c
+++ b/lib/librte_vhost/virtio_net.c
@@ -1056,9 +1056,21 @@ rte_vhost_dequeue_burst(int vid, uint16_t queue_id,
* array, to looks like that guest actually send such packet.
*
* Check user_send_rarp() for more information.
+ *
+ * broadcast_rarp shares a cacheline in the virtio_net structure
+ * with some fields that are accessed during enqueue and
+ * rte_atomic16_cmpset() causes a write if using cmpxchg. This could
+ * result in false sharing between enqueue and dequeue.
+ *
+ * Prevent unnecessary false sharing by reading broadcast_rarp first
+ * and only performing cmpset if the read indicates it is likely to
+ * be set.
*/
- if (unlikely(rte_atomic16_cmpset((volatile uint16_t *)
- &dev->broadcast_rarp.cnt, 1, 0))) {
+
+ if (unlikely(rte_atomic16_read(&dev->broadcast_rarp) &&
+ rte_atomic16_cmpset((volatile uint16_t *)
+ &dev->broadcast_rarp.cnt, 1, 0))) {
+
rarp_mbuf = rte_pktmbuf_alloc(mbuf_pool);
if (rarp_mbuf == NULL) {
RTE_LOG(ERR, VHOST_DATA,
--
1.9.0
next prev parent reply other threads:[~2017-04-07 8:15 UTC|newest]
Thread overview: 49+ messages / expand[flat|nested] mbox.gz Atom feed top
2017-04-07 8:11 [dpdk-stable] patch 'pci: fix device registration on FreeBSD' " Yuanhan Liu
2017-04-07 8:11 ` [dpdk-stable] patch 'kni: fix build with kernel 4.11' " Yuanhan Liu
2017-04-07 8:11 ` [dpdk-stable] patch 'vfio: fix disabling INTx' " Yuanhan Liu
2017-04-07 8:11 ` [dpdk-stable] patch 'vfio: fix secondary process start' " Yuanhan Liu
2017-04-07 8:11 ` [dpdk-stable] patch 'nic_uio: fix device binding at boot' " Yuanhan Liu
2017-04-07 8:11 ` [dpdk-stable] patch 'eal/linux: fix build with glibc 2.25' " Yuanhan Liu
2017-04-07 8:11 ` [dpdk-stable] patch 'examples/ip_fragmentation: fix check of packet type' " Yuanhan Liu
2017-04-07 8:11 ` [dpdk-stable] patch 'kni: fix build on Suse 12 SP3' " Yuanhan Liu
2017-04-10 10:01 ` Nirmoy Das
2017-04-11 12:38 ` Yuanhan Liu
2017-04-07 8:11 ` [dpdk-stable] patch 'net/i40e: fix TC bitmap of VEB' " Yuanhan Liu
2017-04-07 8:11 ` [dpdk-stable] patch 'net/mlx5: fix VLAN stripping indication' " Yuanhan Liu
2017-04-07 8:11 ` [dpdk-stable] patch 'net/i40e: fix compile error' " Yuanhan Liu
2017-04-07 8:11 ` [dpdk-stable] patch 'net/bnx2x: fix transmit queue free threshold' " Yuanhan Liu
2017-04-07 8:11 ` [dpdk-stable] patch 'net/e1000/base: fix multicast setting in VF' " Yuanhan Liu
2017-04-07 8:11 ` [dpdk-stable] patch 'net/mlx5: fix supported packets types' " Yuanhan Liu
2017-04-07 8:11 ` [dpdk-stable] patch 'net/ixgbe/base: fix build error' " Yuanhan Liu
2017-04-07 8:11 ` [dpdk-stable] patch 'net/ixgbe: fix Rx queue blocking issue' " Yuanhan Liu
2017-04-07 8:11 ` [dpdk-stable] patch 'net/ixgbe: fix all queues drop setting of DCB' " Yuanhan Liu
2017-04-07 8:11 ` [dpdk-stable] patch 'net/ixgbe: fix multi-queue mode check in SRIOV mode' " Yuanhan Liu
2017-04-07 8:11 ` [dpdk-stable] patch 'app/testpmd: fix init config for multi-queue " Yuanhan Liu
2017-04-07 8:11 ` [dpdk-stable] patch 'app/testpmd: fix TC mapping in DCB init config' " Yuanhan Liu
2017-04-07 8:11 ` [dpdk-stable] patch 'net/i40e: fix incorrect packet index reference' " Yuanhan Liu
2017-04-07 8:11 ` [dpdk-stable] patch 'net/ixgbevf: set xstats id values' " Yuanhan Liu
2017-04-07 8:11 ` [dpdk-stable] patch 'net/pcap: fix using mbuf after freeing it' " Yuanhan Liu
2017-04-07 8:11 ` [dpdk-stable] patch 'net/ena: fix return of hash control flushing' " Yuanhan Liu
2017-04-07 8:11 ` [dpdk-stable] patch 'net/thunderx: fix 32-bit build' " Yuanhan Liu
2017-04-07 8:11 ` [dpdk-stable] patch 'net/thunderx: fix build on FreeBSD' " Yuanhan Liu
2017-04-07 8:11 ` [dpdk-stable] patch 'net/mlx4: update link status upon probing with LSC' " Yuanhan Liu
2017-04-07 8:11 ` [dpdk-stable] patch 'net/i40e/base: fix potential out of bound array access' " Yuanhan Liu
2017-04-07 8:11 ` [dpdk-stable] patch 'net/mlx5: fix reusing Rx/Tx queues' " Yuanhan Liu
2017-04-07 8:11 ` [dpdk-stable] patch 'net/ixgbe: fix TC bandwidth setting' " Yuanhan Liu
2017-04-07 8:11 ` [dpdk-stable] patch 'net/mlx4: fix returned values upon failed probing' " Yuanhan Liu
2017-04-07 8:11 ` [dpdk-stable] patch 'net/mlx5: " Yuanhan Liu
2017-04-07 8:11 ` [dpdk-stable] patch 'net/fm10k: fix pointer cast' " Yuanhan Liu
2017-04-07 8:11 ` [dpdk-stable] patch 'net/qede: fix missing UDP protocol in RSS offload types' " Yuanhan Liu
2017-04-07 8:11 ` [dpdk-stable] patch 'net/mlx5: fix Tx when first segment size is too short' " Yuanhan Liu
2017-04-07 8:11 ` [dpdk-stable] patch 'vhost: change log levels in client mode' " Yuanhan Liu
2017-04-07 8:11 ` [dpdk-stable] patch 'vhost: fix multiple queue not enabled for old kernels' " Yuanhan Liu
2017-04-07 8:11 ` [dpdk-stable] patch 'vhost: fix max queues' " Yuanhan Liu
2017-04-07 8:11 ` Yuanhan Liu [this message]
2017-04-07 8:11 ` [dpdk-stable] patch 'vhost: fix fd leaks for vhost-user server mode' " Yuanhan Liu
2017-04-07 8:11 ` [dpdk-stable] patch 'net/mlx5: fix an uninitialized variable' " Yuanhan Liu
2017-04-07 8:11 ` [dpdk-stable] patch 'mk: fix shell errors when building with clang' " Yuanhan Liu
2017-04-07 8:12 ` [dpdk-stable] patch 'mk: fix lib filtering when linking app' " Yuanhan Liu
2017-04-07 8:12 ` [dpdk-stable] patch 'examples/quota_watermark: fix requirement for 2M pages' " Yuanhan Liu
2017-04-07 8:12 ` [dpdk-stable] patch 'net/vmxnet3: fix queue size changes' " Yuanhan Liu
2017-04-07 8:12 ` [dpdk-stable] patch 'net/virtio-user: fix overflow' " Yuanhan Liu
2017-04-07 8:12 ` [dpdk-stable] patch 'net/virtio: disable LSC interrupt if MSIX not enabled' " Yuanhan Liu
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=1491552724-3034-39-git-send-email-yuanhan.liu@linux.intel.com \
--to=yuanhan.liu@linux.intel.com \
--cc=ktraynor@redhat.com \
--cc=maxime.coquelin@redhat.com \
--cc=stable@dpdk.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).