From: Dekel Peled <dekelp@mellanox.com>
To: Chao Zhu <chaozhu@linux.vnet.ibm.com>
Cc: "dev@dpdk.org" <dev@dpdk.org>,
"pradeep@us.ibm.com" <pradeep@us.ibm.com>,
Yongseok Koh <yskoh@mellanox.com>, Ori Kam <orika@mellanox.com>,
Thomas Monjalon <thomas@monjalon.net>,
'David Wilder' <dwilder@us.ibm.com>,
Shahaf Shuler <shahafs@mellanox.com>,
Olga Shern <olgas@mellanox.com>
Subject: [dpdk-dev] issue with memory sync on P9
Date: Mon, 18 Mar 2019 09:37:31 +0000 [thread overview]
Message-ID: <VI1PR05MB42244D940D59C5C584C1F65CB6470@VI1PR05MB4224.eurprd05.prod.outlook.com> (raw)
Message-ID: <20190318093731.Y9vYr-LAczeVIuOrsdE2dz6I1A7E8szinXBm0dsy68E@z> (raw)
Hi Chao Zhu,
We have recently encountered runtime issues when running DPDK 18.11 (and later) MLX5 PMD on P9 setup.
Using rte_wmb() was found to be insufficient in specific location, memory sync problem occurred regardless.
Using rte_mb() in this specific location fixed the issue.
--- a/drivers/net/mlx5/mlx5_rxtx.h
+++ b/drivers/net/mlx5/mlx5_rxtx.h
@@ -706,7 +706,7 @@ mlx5_tx_dbrec_cond_wmb(struct mlx5_txq_data *txq, volatile struct mlx5_wqe *wqe,
rte_cio_wmb();
*txq->qp_db = rte_cpu_to_be_32(txq->wqe_ci);
/* Ensure ordering between DB record and BF copy. */
- rte_wmb();
+ rte_mb();
mlx5_uar_write64_relaxed(*src, dst, txq->uar_lock);
if (cond)
rte_wmb();
I see the change in rte_wmb() implementation comes from your patch:
commit d23a6bd04d728226c99d6995c8bd65d49d7e61d1)
Author: Chao Zhu <chaozhu@linux.vnet.ibm.com>
Date: Fri Jul 15 10:30:19 2016 +0800
eal/ppc: fix memory barrier for IBM POWER
I ran some performance tests and got the same results with and without this patch.
Do you have different findings? Will you approve removing this patch?
Regards,
Dekel
> -----Original Message-----
> From: dev <dev-bounces@dpdk.org> On Behalf Of Chao Zhu
> Sent: Wednesday, February 20, 2019 10:07 AM
> To: Thomas Monjalon <thomas@monjalon.net>; 'David Wilder'
> <dwilder@us.ibm.com>
> Cc: dev@dpdk.org; pradeep@us.ibm.com
> Subject: Re: [dpdk-dev] [dpdk-stable] [PATCH v3] mem: Fix anonymous
> mapping on Power9.
>
> +Pradeep
>
> > -----Original Message-----
> > From: Thomas Monjalon <thomas@monjalon.net>
> > Sent: Monday, February 18, 2019 11:47 PM
> > To: David Wilder <dwilder@us.ibm.com>
> > Cc: dev@dpdk.org; chaozhu@linux.vnet.ibm.com
> > Subject: Re: [dpdk-stable] [PATCH v3] mem: Fix anonymous mapping on
> > Power9.
> >
> > Hi,
> >
> > Long time no talk about the IBM Power support status.
> >
> > 16/11/2018 18:35, David Wilder:
> > > --- a/doc/guides/rel_notes/release_18_11.rst
> > > +++ b/doc/guides/rel_notes/release_18_11.rst
> > > +* IBM Power8 is not supported by this release of DPDK. IBM Power9
> > > +is
> > supported.
> >
> > The config file is defconfig_ppc_64-power8-linuxapp-gcc.
> > Given that only Power9 is supported, should we rename it?
> >
> > Is there more changes required to make Power9 working with the latest
> DPDK?
> >
>
>
next reply other threads:[~2019-03-18 9:37 UTC|newest]
Thread overview: 2+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-03-18 9:37 Dekel Peled [this message]
2019-03-18 9:37 ` Dekel Peled
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=VI1PR05MB42244D940D59C5C584C1F65CB6470@VI1PR05MB4224.eurprd05.prod.outlook.com \
--to=dekelp@mellanox.com \
--cc=chaozhu@linux.vnet.ibm.com \
--cc=dev@dpdk.org \
--cc=dwilder@us.ibm.com \
--cc=olgas@mellanox.com \
--cc=orika@mellanox.com \
--cc=pradeep@us.ibm.com \
--cc=shahafs@mellanox.com \
--cc=thomas@monjalon.net \
--cc=yskoh@mellanox.com \
/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).