Hi, The PMD has two paths which can parse the RTE_FLOW items, items, etc.. The path which is returning the error is considered legacy. The other path is initialized depending on a firmware setting. Since the NIC you are using is running an older version, the newer RTE_FLOW offload path is not getting initialized.
I am trying to get the link to the appropriate FW image for you to download. I will share it soon. Thanks Ajit On Tue, Feb 14, 2023 at 10:56 PM Lazuardi Nasution <mrxlazuar...@gmail.com> wrote: > Hi Ajit, > > Is there any update on this? If it is firmware matter, what is suggested > firmware for enabling flow offload with OVN? > > Best regards. > > On Thu, Feb 9, 2023, 12:17 PM Lazuardi Nasution <mrxlazuar...@gmail.com> > wrote: > >> Hi Ajit, >> >> I'm using firmware version 219.0.144.0.of >> >> I'm not sure that the problem is about the capability of the firmware. By >> digging the source code of bnxt PMD, it seems that this problem is related >> to bnxt_validate_and_parse_flow_type() function which throws an error if >> the destination Ethernet address is broadcast Ethernet address. I'm using >> the following URL as reference. >> >> https://github.com/DPDK/dpdk/blob/v21.11/drivers/net/bnxt/bnxt_flow.c#L228 >> >> From what I can understand of David statement, it should not throw an RTE >> error but just leave an incompatible flow non-offloaded. >> >> Best regards. >> >> On Thu, Feb 9, 2023 at 12:14 AM Ajit Khaparde <ajit.khapa...@broadcom.com> >> wrote: >> >>> Hi, >>> From what I can see, it looks like the offload is being attempted on a >>> card which does not have offload functionality enabled. >>> Can you share the FW version on the NICs? >>> >>> If needed, will it be possible for you to update the firmware on the >>> NICs? >>> >>> For the warning regarding flow control setting, let me check and get >>> back. >>> >>> Thanks >>> Ajit >>> >>> On Wed, Feb 8, 2023 at 4:14 AM Lazuardi Nasution <mrxlazuar...@gmail.com> >>> wrote: >>> > >>> > Hi Ajit, >>> > >>> > Have you find the way to overcome this problem? Would you mind to >>> explain why this reserved Ethernet addresses throw error on offloading the >>> flows and not just make related flows non-offloaded? >>> > >>> > Another think, but not so important is bnxt PMD logs warning about >>> cannot do flow control on VF even though I have used none, true or false of >>> interface flow control setting. This warning always appear on OVS >>> restarting. >>> > >>> > Best regards. >>> > >>> > On Tue, Feb 7, 2023, 12:21 AM Lazuardi Nasution < >>> mrxlazuar...@gmail.com> wrote: >>> >> >>> >> Hi Ajit, >>> >> >>> >> I'm using the following versions. >>> >> >>> >> dpdk_version : "DPDK 21.11.2" >>> >> ovs_version : "3.0.1" >>> >> >>> >> Best regards. >>> >> >>> >> On Tue, Feb 7, 2023 at 12:12 AM Ajit Khaparde < >>> ajit.khapa...@broadcom.com> wrote: >>> >>> >>> >>> On Mon, Feb 6, 2023 at 9:02 AM Lazuardi Nasution < >>> mrxlazuar...@gmail.com> wrote: >>> >>> > >>> >>> > Hi David, >>> >>> > >>> >>> > I think I can understand your opinion. So your target is to >>> prevent frames with those ethernet addresses from reaching CP, right? FYI, >>> I'm using bonded VFs of bonded PFs as OVS-DPDK interfaces, so offcourse >>> LACP should be handled by bonded PFs only. >>> >>> What is the version of DPDK & OVS used here, BTW? Thanks >>> >>> >>> >>> > >>> >>> > Best regards, >>> >>> > >>> >>> > On Mon, Feb 6, 2023 at 11:54 PM David Marchand < >>> david.march...@redhat.com> wrote: >>> >>> >> >>> >>> >> On Mon, Feb 6, 2023 at 5:46 PM Lazuardi Nasution < >>> mrxlazuar...@gmail.com> wrote: >>> >>> >> > >>> >>> >> > HI David, >>> >>> >> > >>> >>> >> > Don't you think that offload of reserved Ethernet address >>> should be disabled by default? >>> >>> >> >>> >>> >> What OVN requests in this trace (dropping) makes sense to me if >>> those >>> >>> >> lacp frames are to be ignored at the CP level. >>> >>> >> I don't see why some ethernet address would require some special >>> >>> >> offloading considerations, but maybe others have a better opinion >>> on >>> >>> >> this topic. >>> >>> >> >>> >>> >> >>> >>> >> -- >>> >>> >> David Marchand >>> >>> >> >>> >>
smime.p7s
Description: S/MIME Cryptographic Signature
_______________________________________________ discuss mailing list disc...@openvswitch.org https://mail.openvswitch.org/mailman/listinfo/ovs-discuss