Replies: 12 comments 39 replies
-
I will start out by saying that I am new to magma and have not used Sercomm. However, when connecting a Pixel 6 to srsRAN gnb (n78) and Open5GS core I was experiencing similar behavior (UE drops the connection). To resolve the issue I updated the phone's software to
Hope that helps. |
Beta Was this translation helpful? Give feedback.
-
For debugging this issue, could you please share the debug logs (UE, gnb and AMF) |
Beta Was this translation helpful? Give feedback.
-
Is SUCI enabled by default in Magma? Can you please share the pcap/snapshot of Registration Accept specifically the 5GS Network Feature Support IE. The 5G SA in 1.8 was tested with Huawei Mate 30 UE |
Beta Was this translation helpful? Give feedback.
-
@panyogesh @prabinakpattnaik Adding logs for the attach/release cycle. Except for the gNB pcap, they all cover the same attach/release even though the timestamps don't match. In order: AMF/MME log from AGW, logcat from the UE, pcapng from the AGW, gNB text log. The last file is a pcap for the gNB for the same problem but at a different time. I can only get this pcap by getting Sercomm to provide it so its harder to capture. |
Beta Was this translation helpful? Give feedback.
-
I am wondering if the 5G container images I'm using are older than I should use (00787b2 from January 2023). I'm trying a later version although I'm not sure which ones since January are relatively stable. |
Beta Was this translation helpful? Give feedback.
-
This issue (UE disconnect after 10 seconds) persists. I've now reproduced on several different commits on 1.9 including the December 2023 commit. I've also rebuilt for ARM servers and still there. Will keep working it but would love more suggestions on what to try. |
Beta Was this translation helpful? Give feedback.
-
Yogesh and Jim and their teams have been working out a path to get this to work. Possible 1.9.2 inclusion. |
Beta Was this translation helpful? Give feedback.
-
IMO this discussion should turn into an issue, and probably ultimately into a PR of some kind. |
Beta Was this translation helpful? Give feedback.
-
An update to cover the last two months of offline work with Yogesh and team. TL;DR -- getting one known good 5G UE to work and identifying other working UEs is still an issue. To recap, I'm using a recent pre-1.9 commit ( For much of the above thread, I could get a Pixel 7 Pro to connect for ~10 seconds but then it disconnected with Some changes enroute:
Rebuild the containers and redeploy the AGW:
On the Pixel 7 Pro, at phone keypad enter:
Which brings up a set of options. Change:
While using Wireshark to debug, it was helpful to turn on:
None of these changes stopped the disconnect problem. Given I was stuck with magma, I deployed Open5GS to see if that had similar issues. With Open5GS, it stayed connected. This may be because of repeated DNS queries by the UE that kept the connection up. Not sure why those DNS requests didn't happen with magma. Now, it gets interesting. Last week, I updated the UE to Android 14. At that point, NGAP connection attempts stopped appearing between the gNB and AGW or Open5GS -- a step backwards. I then downgrade to Android 13 and, even though the same UE, same OS, same SIM, the configuration that at least connected temporarily before no longer tries to connect (at least as viewed from 5G core). I parallel with this I also tried other types of UE. A Pixel 6 Pro, OnePlus 9, and a Global Telecom Titan 5000 5G Access Point. They exhibit the same behavior as the Pixel 7 Pro now does -- no connection attempt. I also tried all of these same UE using an active T-Mobile SIM. With the exception of the OnePlus 9, they all successfully connect to T-Mobile. The Pixel 7 Pro and Pixel 6 Pro also connect to our 4G LTE Magma network. The Access Point is 5G only so doesn't connect to the LTE network. The OnePlus 9 doesn't connect to anything. In summary, the only components in common across all of these 5G test scenarios are the gNB and the SIM. The Pixel 7 Pro sort of worked for a while but now doesn't. The gNB wasn't impacted by the Android upgrade/downgrade so no reason to think that would be responsible. APNs were configured across all UE, magma, and Open5GS. PLMN=00101. The SIM is from sysmocom and was programmed with pySim. There was no specific 5G info included in the SIM programming. That's where I am. Thanks to Yogesh & Co. for all their help. I'm attaching a couple of recent pcap files for the disconnect issue. I'm also including an adb logcat file for the Pixel 7 Pro during the period I would expect it to connect. Any help would be appreciated. |
Beta Was this translation helpful? Give feedback.
-
ue-no-disconnect-open5gs-2024-02-29-1.pcapng; is doing a GUTI registration, you can see the Ue security capability listing EEA and EIA. the 5Gs core then responds with InitialContextSetupRequest including support for EEA and EIA, the magma core in 'ue-connect-release-while-pinging-2023-02-16.pcapng' has those as unsupported fields. The open5Gs responds to the InitialContextSetupRequest with the DownlinkNASTransport encrypted. there is no such signaling from the magma core. I would assume if you enable debugging on everything you would see that magma is not able to respond to the request without the encyption. |
Beta Was this translation helpful? Give feedback.
-
Thank you. I will try.
…On Tue, Apr 2, 2024 at 11:11 PM Mark Harris ***@***.***> wrote:
ue-no-disconnect-open5gs-2024-02-29-1.pcapng; is doing a GUTI
registration, you can see the Ue security capability listing EEA and EIA.
the 5Gs core then responds with InitialContextSetupRequest including
support for EEA and EIA, the magma core in
'ue-connect-release-while-pinging-2023-02-16.pcapng' has those as
unsupported fields. The open5Gs responds to the InitialContextSetupRequest
with the DownlinkNASTransport encrypted. there is no such signaling from
the magma core. I would assume if you enable debugging on everything you
would see that magma is not able to respond to the request without the
encyption.
I know for the iPhone SUCI and subsequent encryption requirement we have
set;
/etc/magma/templates/mme.conf.template
NAS Change line
ORDERED_SUPPORTED_CIPHERING_ALGORITHM_LIST = [ "EEA0" , "EEA1" ,
"EEA2" ];
To
ORDERED_SUPPORTED_CIPHERING_ALGORITHM_LIST = [ "EEA2" , "EEA1" , "EEA0" ];
then stop and restart magma services.
—
Reply to this email directly, view it on GitHub
<#15297 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/ABTSNMFPLNE3ENXB4KLLMZDY3N6PJAVCNFSM6AAAAAA4E22I2SVHI2DSMVQWIX3LMV43SRDJONRXK43TNFXW4Q3PNVWWK3TUHM4DSOJRGU3TI>
.
You are receiving this because you were mentioned.Message ID:
***@***.***>
--
Jim Blakley
Living Edge Lab Associate Director
Carnegie Mellon University
|
Beta Was this translation helpful? Give feedback.
-
Beta Was this translation helpful? Give feedback.
-
We are building out our 5G test environment as an evolution of our current LTE network. We're on CBRS n48 for 5G. The AGW, gNB, and Orc8r are up, connected and enabled for 5G. I have tried to connect 2 different 5G UE with marginal success. Progress so far:
UE 1 -- Pixel 7 Pro. We've been using this already in our LTE network and it supports 5G SA on CBRS n48. Initially, it wouldn't even see our 5G network but, based on info from our gNB vendor (Sercomm), when I moved the network to PLMN 00101, it did see it and completes registration. However, after (reliably) ~10 seconds, the UE drops the connection. While the
InitialContextSetupResponse
comes from gNB to AMF, theNAS Registration Complete
message from UE to AMF never arrives. AMF/MME log does say registration complete but the UE logs never go to CONNECTED. A potentially related thing -- when I scan in our lab, there are other networks using the 00101 PLMN -- one on band L2 and one on band L66. They are not ours but could be confusing our UE. At a dead end now, so any ideas on this?UE 2 -- Oneplus 9 Pro. This model is known good for 5G SA on n48 but our two UEs of this type don't even see our network. Have tried both on the 00101 PLMN and our standard network PLMN. Although they were sold as unlocked, they may only be unlocked for T-Mobile, AT&T, and Verizon. Is there a reliable way to see whether they are locked such that they won't connect to a private network? Even better, any one know a way to carrier unlock them? Any other explanations for this behavior.
Happy to share additional info as needed.
Now to a more general question -- if I can't get any of these UEs to work, I'll have to find another. Has anyone found and connected to Magma a 5G SA smartphone UE on band n48 -- if so what type and how did you verify it was unlocked before you got it? If you've done it, just not on n48, please share that too!
@jordanvrtanoski
Beta Was this translation helpful? Give feedback.
All reactions