[dpdk-dev] mbuf->packet_type

[dpdk-dev] mbuf->packet_type

[Levend Sayar]

Hi all.

Recently I encountered a dont-know-why issue about mbuf->packet_type.

Two different setups behave differently.

One is a bare metal server, one KVM based OpenStack VM (virtio).
Both use the uio driver. Namely only hardware is different. Software stack
is the same.

But this macro does not work on OpenStack platform.

*RTE_ETH_IS_IPV4_HDR(mbuf->packet_type))*

where

*#define  RTE_ETH_IS_IPV4_HDR(ptype) ((ptype) & RTE_PTYPE_L3_IPV4)*

I manually coded to check IPv4 like this:




*struct rte_ether_hdr* ether_hdr = rte_pktmbuf_mtod(mbuf, struct
rte_ether_hdr*);uint16_t ether_type = N2H16(ether_hdr->ether_type);if
(likely(ether_type == RTE_ETHER_TYPE_IPV4)) {*

Do you have any idea about this?
I suppose whoever fills that mbuf data packet_type field is not working in
a virtual environment.
Which entity fills that fields? Hardware driver, uio, dpdk?

Best,
Levend

Re: [dpdk-dev] mbuf->packet_type

[Levend Sayar]

Hi all.

Nobody cared about my mail. Am I in the wrong place?
If so, can you forward me to the right place please?

Best,
Levend

On Mon, Nov 2, 2020 at 1:45 PM Levend Sayar <levendsayar@gmail.com> wrote:

> Hi all.
>
> Recently I encountered a dont-know-why issue about mbuf->packet_type.
>
> Two different setups behave differently.
>
> One is a bare metal server, one KVM based OpenStack VM (virtio).
> Both use the uio driver. Namely only hardware is different. Software stack
> is the same.
>
> But this macro does not work on OpenStack platform.
>
> *RTE_ETH_IS_IPV4_HDR(mbuf->packet_type))*
>
> where
>
> *#define  RTE_ETH_IS_IPV4_HDR(ptype) ((ptype) & RTE_PTYPE_L3_IPV4)*
>
> I manually coded to check IPv4 like this:
>
>
>
>
> *struct rte_ether_hdr* ether_hdr = rte_pktmbuf_mtod(mbuf, struct
> rte_ether_hdr*);uint16_t ether_type = N2H16(ether_hdr->ether_type);if
> (likely(ether_type == RTE_ETHER_TYPE_IPV4)) {*
>
> Do you have any idea about this?
> I suppose whoever fills that mbuf data packet_type field is not working in
> a virtual environment.
> Which entity fills that fields? Hardware driver, uio, dpdk?
>
> Best,
> Levend
>

Re: [dpdk-dev] mbuf->packet_type

[Juhamatti Kuusisaari]

Hello Levend,

The packet type is indeed filled by hw in case there is a support for
it. I think your application may be missing a ptype handling function
which checks the supported ptype handling of the hw and registers a
callback handling in case the hw does not support it.

Please see more details e.g. in l3fwd example application about the
ptype checks and callback registration.

Best regards,
--
 Juhamatti

On Tue, 3 Nov 2020 at 09:33, Levend Sayar <levendsayar@gmail.com> wrote:
>
> Hi all.
>
> Nobody cared about my mail. Am I in the wrong place?
> If so, can you forward me to the right place please?
>
> Best,
> Levend
>
> On Mon, Nov 2, 2020 at 1:45 PM Levend Sayar <levendsayar@gmail.com> wrote:
>
> > Hi all.
> >
> > Recently I encountered a dont-know-why issue about mbuf->packet_type.
> >
> > Two different setups behave differently.
> >
> > One is a bare metal server, one KVM based OpenStack VM (virtio).
> > Both use the uio driver. Namely only hardware is different. Software stack
> > is the same.
> >
> > But this macro does not work on OpenStack platform.
> >
> > *RTE_ETH_IS_IPV4_HDR(mbuf->packet_type))*
> >
> > where
> >
> > *#define  RTE_ETH_IS_IPV4_HDR(ptype) ((ptype) & RTE_PTYPE_L3_IPV4)*
> >
> > I manually coded to check IPv4 like this:
> >
> >
> >
> >
> > *struct rte_ether_hdr* ether_hdr = rte_pktmbuf_mtod(mbuf, struct
> > rte_ether_hdr*);uint16_t ether_type = N2H16(ether_hdr->ether_type);if
> > (likely(ether_type == RTE_ETHER_TYPE_IPV4)) {*
> >
> > Do you have any idea about this?
> > I suppose whoever fills that mbuf data packet_type field is not working in
> > a virtual environment.
> > Which entity fills that fields? Hardware driver, uio, dpdk?
> >
> > Best,
> > Levend
> >