Strange behavior of ECN in RLC.

[Maria Eduarda Veras Martins]

Hello everyone,

I need to use the ECN field in RLC-UM and I noticed that all packets have the ECN Not-ECT. I am using two applications in my simulation, one with DCTCP (I modified it to use ECN ECT1 at initialization) and another with Cubic. I analyzed the pcaps and the packets appear with the correct ECN. I also checked the PDCP and saw that the ECN is correct there, only in the RLC it becomes 0 for some reason. Does anyone know the reason or have experienced something similar?

I am obtaining the ECN as follows:


(I am not sure if I should post here or on ns3-users. If there is any issue, I can delete it later.)

Kind Regards,
Maria.

[Kent Huns]

Hi Maria

Maybe ns-3/5G-LENA cannot process RLC-UM correctly.

For example,  UM's header mustn't' contain a D/C bit but always checked. (like your log) 

So I'm not sure if the other bits are correctly read as well.

https://www.sharetechnote.com/html/5G/5G_RLC.html

I've experienced similar problem in receiving process.

https://gitlab.com/cttc-lena/nr/-/issues/123

Thanks,

Kent

[Maria Eduarda Veras Martins]

Hi Kent!

I saw in the issue that apparently RLC-AM is working fine, but when I tried to retrieve the ECN it still exhibited the same behavior.

Do you know anything about it?

Thank you,

Maria.

--
You received this message because you are subscribed to the Google Groups "5G-LENA-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to 5g-lena-user...@googlegroups.com.
To view this discussion on the web, visit https://groups.google.com/d/msgid/5g-lena-users/a08494a7-0dfd-497d-aa59-0e285f3547c0n%40googlegroups.com.

[Kent Huns]

You can check if the 0 value is truly "numeric 0" or "null".   ....sorry, I have no more ideas. I don't know at all about the upper layer.

[Maria Eduarda Veras Martins]

The "error" is actually in the PDCP layer. If you try to access the IPv4 header after adding the PDCP header, the IP header always remains null. Is this expected behavior?

[Tom Henderson]

Yes, you can only Peek a specific header if there are no other headers
in front of it. If you need to access a header that is not at the head
of the packet, you need to remove those prepended header(s) first.

You can possibly work around this by tagging the packet (ns-3 Packet
tag) and performing the actual mark when it leaves the other
NrNetDevice. For instance, when entering the NrNetDevice, define a new
Packet tag type that copies the ECN byte into it. When you need to
inspect at the RLC layer after the PDCP header has been prepended, use
the Packet API to find the tag and read it. If needed to mark the
packet (from ECT-1 to CE), call Packet::ReplacePacketTag() to update the
value. Then, just before the packet leaves the egress NrNetDevice,
fetch the tag again and ensure that the ECN value is updated if needed.

On 10/3/24 09:10, Maria Eduarda Veras Martins wrote:
> The "error" is actually in the PDCP layer. If you try to access the IPv4
> header after adding the PDCP header, the IP header always remains null.
> Is this expected behavior?

> Screenshot from 2024-10-03 13-03-25.png
>
>
> Em qui., 3 de out. de 2024 às 11:14, Kent Huns

> <kent.sa...@gmail.com <mailto:kent.sa...@gmail.com>> escreveu:

>
> You can check if the 0 value is truly "numeric 0" or "null".
>  ....sorry, I have no more ideas. I don't know at all about the
> upper layer.
>
> On Thu, Oct 3, 2024 at 10:17 PM Maria Eduarda Veras Martins
> <me...@cin.ufpe.br <mailto:me...@cin.ufpe.br>> wrote:
>
> Hi Kent!
>

> I saw in the issue that *apparently* RLC-AM is working fine, but

> when I tried to retrieve the ECN it still exhibited the same
> behavior.

> Screenshot from 2024-10-03 10-05-27.png
> Do you know anything about it?
>
> Thank you,
> Maria.
>
> Em qui., 3 de out. de 2024 às 09:41, Kent Huns

> <kent.sa...@gmail.com <mailto:kent.sa...@gmail.com>>

> Captura de tela 2024-10-02 151249.png
>
> I am obtaining the ECN as follows:

> Captura de tela 2024-10-02 151343.png
>
> (I am not sure if I should post here or on ns3-users. If
> there is any issue, I can delete it later.)
>
> Kind Regards,
> Maria.
>
> --
> You received this message because you are subscribed to the
> Google Groups "5G-LENA-users" group.
> To unsubscribe from this group and stop receiving emails
> from it, send an email to
> 5g-lena-user...@googlegroups.com

> <mailto:5g-lena-user...@googlegroups.com>.

> To view this discussion on the web, visit

> https://groups.google.com/d/msgid/5g-lena-users/a08494a7-0dfd-497d-aa59-0e285f3547c0n%40googlegroups.com <https://groups.google.com/d/msgid/5g-lena-users/a08494a7-0dfd-497d-aa59-0e285f3547c0n%40googlegroups.com?utm_medium=email&utm_source=footer>.

>
> --
> You received this message because you are subscribed to the Google
> Groups "5G-LENA-users" group.
> To unsubscribe from this group and stop receiving emails from it, send
> an email to 5g-lena-user...@googlegroups.com

> <mailto:5g-lena-user...@googlegroups.com>.

> To view this discussion on the web, visit

> https://groups.google.com/d/msgid/5g-lena-users/CAEdnndnqdJ6O901VUEr8zwf%2BG91KPdHSR5Q_EJmyvjTdSaEibw%40mail.gmail.com <https://groups.google.com/d/msgid/5g-lena-users/CAEdnndnqdJ6O901VUEr8zwf%2BG91KPdHSR5Q_EJmyvjTdSaEibw%40mail.gmail.com?utm_medium=email&utm_source=footer>.

[Maria Eduarda Veras Martins]

I see, I didn't know about this in peek function.

I just don't understand why NrNetDevice; do I need to do this before sending the packet? Would it be an issue to add the tag only in the PDCP?

Thank you,
Maria.

To unsubscribe from this group and stop receiving emails from it, send an email to 5g-lena-user...@googlegroups.com.
To view this discussion on the web, visit https://groups.google.com/d/msgid/5g-lena-users/308a9fcf-73b1-6e4f-385b-83f892d66b4d%40tomh.org.

[Tom Henderson]

On 10/3/24 11:06, Maria Eduarda Veras Martins wrote:
> I see, I didn't know about this in peek function.
> I just don't understand why NrNetDevice; do I need to do this before
> sending the packet? Would it be an issue to add the tag only in the PDCP?

You can add it wherever you have access to the IP header.

Note that if the SDAP layer is added (I am working on this for
sidelink), the SDAP header will be added above the PDCP layer; hence, it
would be more future proof if this tag were just added in the body of
NrNetDevice::Send() and SendFrom() methods.

- Tom