DPDK usage discussions
 help / color / mirror / Atom feed
* [dpdk-users] binding PF and VF to DPDK
@ 2019-01-28  3:26 Jasmine Sanghvi
  0 siblings, 0 replies; only message in thread
From: Jasmine Sanghvi @ 2019-01-28  3:26 UTC (permalink / raw)
  To: users

Hi:


Our environment:

DPDK Version: 18.02
Kernel: 4.14.56
NIC: XL710/i40e/i40evf
SRIOV (linux i40e driver version 2.7.11 / i40ef driver version 3.6.10)

On host, we create n VFs on the physical device, by setting
/sys/bus/pci/device/$PF/sriov_numvfs.

Our dpdk application that binds to these VFs and does some packet processing.
PF stays bound to Linux i40e driver. We are not using the Linux PF interface
for any other purpose.
Under continuous heavy load we find that our application stops receiving
traffic after a while.
dmesg displays a msg to the tune of "TX driver issue detected, PF reset issued".

Looking at the i40e linux driver code, it appears that i40e is detecting "Malicious
driver".  I don't see any way to disable "Malicious Driver Detection"
in i40e driver (no module parameter).

Alternative was to bind PF also to dpdk (igb_uio) and not start PF eth dev.
However when I bind PF to dpdk, all VFs disappear.

I tried binding PF before and after...same result.  After I bind PF to dpdk,
rte_eth_dev_count() only reflects PF and cannot get dev_info corresponding to
VFs.

My question is can PF and VFs be both bound to DPDK?


Thanks,

Jasmine

^ permalink raw reply	[flat|nested] only message in thread

only message in thread, other threads:[~2019-01-28  3:26 UTC | newest]

Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2019-01-28  3:26 [dpdk-users] binding PF and VF to DPDK Jasmine Sanghvi

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).