From: "Hirok J. Bora" <hirok.gen@gmail.com>
To: Suchetha p <suchetha2005@gmail.com>
Cc: users <users@dpdk.org>
Subject: Re: [dpdk-users] Generic Virtualization approach of DPDK based application using SR-IOV and PCI Passthrough to be compatible with all Hardware NICs
Date: Sat, 1 Feb 2020 23:12:26 +0530 [thread overview]
Message-ID: <CABMKPHWgH5+hkrB25giG9RExVDuH50n=mL_-TAR5PzTQRAPx7Q@mail.gmail.com> (raw)
In-Reply-To: <CAOGjjANzC=yFTsesDH1zTam56cUpy1e4PnsE6Ei1hdUotzbsrw@mail.gmail.com>
Avoid SR-IOV and implement OVS on the host. Implement acceleration at host
level first then guest level, if needed.
SR-IOV does not provide L2 level switching flexibility. Do use overlay
between ovs and the guest, implement vxlan. This will give you 100% control
on the HW. All the best 👍
On Sat, Feb 1, 2020, 10:50 PM Suchetha p <suchetha2005@gmail.com> wrote:
> Hi,
>
> We are trying to bring up KVM based VMs on HP Gen10 Servers running on RHEL
> 7.6 and having Intel X710 NIC's.
>
> The VMs are also running on RHEL 7.6 and the DPDK version used in the guest
> application is 18.08.
>
> Intel DPDK VF drivers (igb_uio) compiled and loaded currently in our VM are
> not compatible with the Host PF drivers(vfio-pci) for Intel X710 Ethernet
> NICs.
>
> The igb_uio driver may be only compatible with Intel NICs like 82599 as it
> worked in our lab for HP Gen9 servers running RHEL 7.6 having Intel 82599
> NICs.
>
> Issues observed with DPDK VF driver igb_uio for Intel X710 NICs:
>
> The outgoing packets from DPDK application are leaving VM via VF towards
> Host’s Physical ports and reaching remote Host’s Physical ports.
>
> But the incoming packets from Host’s Physical port are not reaching VM via
> VF.
>
> Could you please confirm whether igb_uio driver is compatible with only
> Intel NICs like 82599 ?
>
> So we had to change PCI device binding to another driver vfio-pci available
> in RHEL7.
>
> But based on the debugging of our DPDK based application with VF driver
> vfio-pci on X710 NICs we observed that the incoming packets are VLAN
> stripped by HW when the OFFLOAD flags are not set for the same.
>
> why does X710 NIC VF driver remove the VLAN without RX offload VLAN strip
> flags set?
>
> Could you please check and confirm if the unexpected VLAN removal behaviour
> of X710 NIC VF driver vfio-pci is a known bug?
>
> We want to generalize DPDK implementation for all HW NICs.Could you please
> clarify if this is achievable with igb_uio or vfio-pci driver ?
>
> We would appreciate if you could provide your feedback and suggestions
> towards generic solution.
>
> Thanks,
>
> Suchetha
>
next prev parent reply other threads:[~2020-02-01 17:42 UTC|newest]
Thread overview: 3+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-02-01 17:20 Suchetha p
2020-02-01 17:42 ` Hirok J. Bora [this message]
2020-02-06 13:19 ` madhukar mythri
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='CABMKPHWgH5+hkrB25giG9RExVDuH50n=mL_-TAR5PzTQRAPx7Q@mail.gmail.com' \
--to=hirok.gen@gmail.com \
--cc=suchetha2005@gmail.com \
--cc=users@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).