Re: [dpdk-dev] l3fwd can't launch on VF when we use i40e

Re: [dpdk-dev] l3fwd can't launch on VF when we use i40e

[Xu, Qian Q]

+ dpdk.org mailing list. I also thought the patch to check port CRC strip then make the port fail to start is not proper.
For testpmd, we can use -crc-strip, but for l3fwd, we may need change code to enable crc-strip, and for all application, we need change the code; this is not acceptable.
Users may think it's a bug of R16.07 since R16.04 didn't check that. From my view, crc-strip is not a "MUST" check for the port start. We can configure the value after
the port start.
Any thoughts?

From: Yao, Lei A
Sent: Friday, July 22, 2016 9:36 AM
To: Topel, Bjorn <bjorn.topel@intel.com>
Cc: Xu, Qian Q <qian.q.xu@intel.com>
Subject: l3fwd can't launch on VF when we use i40e

Hi, Bjorn

This is lei who is running DPDK testing on VMware with 16.07 RC3. Now I met one issue, after I enable SR-IOV on VMware with i40e related NIC, I can't launch l3fwd example on VF because of this error info:
i40evf_dev_configure(): VF can't disable HW CRC Strip.

I see that you have submit following patch related to this scenario
commit 1bbcc5d21129168a212e7d755751b0d4742d20d9
Author: Björn Töpel <bjorn.topel@intel.com>
Date:   Fri Apr 22 07:39:22 2016 +0200

    i40evf: report error for unsupported CRC stripping config

    On hosts running a non-DPDK PF driver, the VF has no means of changing
    the HW CRC strip setting for a RX queue. It's implicitly enabled.

    This patch checks if the host is running a non-DPDK PF kernel driver,
    and returns an error, if HW CRC stripping was not requested in the port
    configuration.

    Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
    Acked-by: Helin Zhang <helin.zhang@intel.com>

Do you know is there any way I can bypass this when I use l3fwd? I know when we use testpmd, we can add --crc-strip enable to bypass this blocker. But I don't know the method on l3fwd. Thanks.

Best Regards
Lei

Re: [dpdk-dev] l3fwd can't launch on VF when we use i40e

[Björn Töpel]

Firstly, sorry about the late reply.

> From my view, crc-strip is not a "MUST" check for the port start. We can configure the value after
> the port start.
> Any thoughts?

>From my perspective, a non-supported configuration should return an
error code to the API user.

> Do you know is there any way I can bypass this when I use l3fwd?

One work-around is enabling .hw_strip_crc for i40 VFs at runtime, by
e.g. using rte_eth_dev_info_get() to determine what driver that's in
use. That, however, does require a l3fwd patch.


Björn