feat: add F-TEID allocation support #840

gruyaume · 2024-08-05T18:55:42Z

Description

As of 3GPP Release 16, F-TEID allocation is done in the UPF instead of the SMF. This PR implements F-TEID allocation as well as the necessary PFCP communication changes.

Summary

The UPF will set the FTUP feature flag in its PFCP Association Setup Response to the SMF
The SMF must set the CHOOSE flag in its request to establishment a new PDR in the PFCP Session Establishment Request
When this CHOOSE flag is set, the UPF allocates a F-TEID using a TEID generator
The UPF sets the F-TEID in its PFCP Session Establishment Response

Changes in the SMF and UPF are intertwined

This change and the SMF are intertwined and have to be merged close together so that the network keeps on working on main:

feat: request UPF to allocate F-TEID smf#334

Reference

3GPP Release 16

Relevant information from the 3GPP Release 16 29.244 document:

F-TEID shall be only allocated by the UPF. The UPF shall set the FTUP feature flag in the UP Function Features IE and the SMF shall request the UPF to allocate the F-TEID. The UPF shall reject a request to establish a new PDR with
an F-TEID allocation in the SMF option, with the cause "Invalid F-TEID allocation option". As an exception, the
UPF shall accept the PFCP Session Establishment Request message with a PDR including an existing F-TEID
to re-establish a PFCP session during a restoration procedure.

The SMF shall request the UPF to allocate the F-TEID by setting the CHOOSE flag in the Local F-TEID. And if the PDR(s) is created successfully, the UPF shall return the F-TEID(s) it has assigned to the PDR(s) in the PFCP Session Establishment Response or PFCP Session Modification Response.

3GPP Release 16: Interface between the Control Plane and the User Plane nodes

omecproject · 2024-08-05T18:55:46Z

Can one of the admins verify this patch?

gab-arrobo · 2024-10-17T15:38:08Z

@gruyaume please rebase your PR.

gab-arrobo · 2024-10-17T18:58:55Z

@gruyaume,
FYI, the 2 GHAs that fail are because of the same reason:

=== RUN   TestUPFBasedUeIPAllocation
    basic_test.go:102: 
        	Error Trace:	/home/runner/work/upf/upf/test/integration/basic_test.go:102
        	Error:      	"[0xc000de20c0 0xc000de2180]" should have 1 item(s), but has 2
        	Test:       	TestUPFBasedUeIPAllocation
--- FAIL: TestUPFBasedUeIPAllocation (5.18s)

gruyaume · 2024-10-17T19:16:05Z

@gruyaume, FYI, the 2 GHAs that fail are because of the same reason:

=== RUN   TestUPFBasedUeIPAllocation
    basic_test.go:102: 
        	Error Trace:	/home/runner/work/upf/upf/test/integration/basic_test.go:102
        	Error:      	"[0xc000de20c0 0xc000de2180]" should have 1 item(s), but has 2
        	Test:       	TestUPFBasedUeIPAllocation
--- FAIL: TestUPFBasedUeIPAllocation (5.18s)

Yeah this is still a draft, I'm trying to figure out why the Created PDR's aren't returned in the PFCP Session Establishment Response.

gab-arrobo

Overall, it looks good to me. However, when running an E2E test using this PR along with SMF #334, the test fails, as shown below. Did you get a chance to run an E2E test with these changes?

TASK [simulator : debug] ***********************************************************************************************
fatal: [node1]: FAILED! => {
    "msg": [
        "Profile Name: profile2, Profile Type: pdusessest",
        "Ue's Passed: 0, Ue's Failed: 1",
        "Profile Errors:",
        "imsi:imsi-208930100007510, profile timeout",
        "Profile Status: FAIL"
    ]
}

docs/configuration-guide.md

pfcpiface/fteid.go

gruyaume · 2024-11-13T12:24:08Z

Overall, it looks good to me. However, when running an E2E test using this PR along with SMF #334, the test fails, as shown below. Did you get a chance to run an E2E test with these changes?
TASK [simulator : debug] ***********************************************************************************************
fatal: [node1]: FAILED! => {
    "msg": [
        "Profile Name: profile2, Profile Type: pdusessest",
        "Ue's Passed: 0, Ue's Failed: 1",
        "Profile Errors:",
        "imsi:imsi-208930100007510, profile timeout",
        "Profile Status: FAIL"
    ]
}

Yes I've tested this end to end and it worked fine with all the 5 profiles I tested with. I'll try again to make sure.

gruyaume · 2024-11-13T13:15:53Z

Overall, it looks good to me. However, when running an E2E test using this PR along with SMF #334, the test fails, as shown below. Did you get a chance to run an E2E test with these changes?
TASK [simulator : debug] ***********************************************************************************************
fatal: [node1]: FAILED! => {
    "msg": [
        "Profile Name: profile2, Profile Type: pdusessest",
        "Ue's Passed: 0, Ue's Failed: 1",
        "Profile Errors:",
        "imsi:imsi-208930100007510, profile timeout",
        "Profile Status: FAIL"
    ]
}
Yes I've tested this end to end and it worked fine with all the 5 profiles I tested with. I'll try again to make sure.

I tested it again and it still works on my side, I attached the gnbsim logs.
gnbsim_logs.txt

Let me know if you have more information about failures. Logs and/or captures would be useful.

gab-arrobo · 2024-11-13T18:11:55Z

I tested it again and it still works on my side, I attached the gnbsim logs. gnbsim_logs.txt

Let me know if you have more information about failures. Logs and/or captures would be useful.

My E2E tests still fail, except Profile 1 (Registration request: no UP involvement). Let me take another look into the PR and will share logs later today/tomorrow. Also, I asked @sureshmarikkannu to test these PRs

        "Profile Name: profile1, Profile Type: register",
        "Ue's Passed: 5, Ue's Failed: 0",
        "Profile Status: PASS",
        "Profile Name: profile5, Profile Type: deregister",
        "Ue's Passed: 0, Ue's Failed: 5",
        "Profile Errors:",
        "imsi:imsi-208930100007543, profile timeout",
        "imsi:imsi-208930100007544, profile timeout",
        "imsi:imsi-208930100007542, profile timeout",
        "imsi:imsi-208930100007540, profile timeout",
        "imsi:imsi-208930100007541, profile timeout",
        "Profile Status: FAIL",
        "Profile Name: profile3, Profile Type: anrelease",
        "Ue's Passed: 0, Ue's Failed: 5",
        "Profile Errors:",
        "imsi:imsi-208930100007524, profile timeout",
        "imsi:imsi-208930100007523, profile timeout",
        "imsi:imsi-208930100007522, profile timeout",
        "imsi:imsi-208930100007521, profile timeout",
        "imsi:imsi-208930100007520, profile timeout",
        "Profile Status: FAIL",
        "Profile Name: profile7, Profile Type: uereqpdusessrelease",
        "Ue's Passed: 0, Ue's Failed: 5",
        "Profile Errors:",
        "imsi:imsi-208930100007562, profile timeout",
        "imsi:imsi-208930100007564, profile timeout",
        "imsi:imsi-208930100007561, profile timeout",
        "imsi:imsi-208930100007563, profile timeout",
        "imsi:imsi-208930100007560, profile timeout",
        "Profile Status: FAIL",
        "Profile Name: profile2, Profile Type: pdusessest",
        "Ue's Passed: 0, Ue's Failed: 5",
        "Profile Errors:",
        "imsi:imsi-208930100007513, profile timeout",
        "imsi:imsi-208930100007512, profile timeout",
        "imsi:imsi-208930100007511, profile timeout",
        "imsi:imsi-208930100007514, profile timeout",
        "imsi:imsi-208930100007510, profile timeout",
        "Profile Status: FAIL",
        "Profile Name: profile4, Profile Type: uetriggservicereq",
        "Ue's Passed: 0, Ue's Failed: 5",
        "Profile Errors:",
        "imsi:imsi-208930100007534, profile timeout",
        "imsi:imsi-208930100007532, profile timeout",
        "imsi:imsi-208930100007530, profile timeout",
        "imsi:imsi-208930100007533, profile timeout",
        "imsi:imsi-208930100007531, profile timeout",
        "Profile Status: FAIL"

gruyaume · 2024-11-15T11:52:24Z

today

@gab-arrobo did you get time to find more details about the problem you are experimenting? Any help here would be useful in driving this one to completion.

gab-arrobo · 2024-11-15T17:45:49Z

today

@gab-arrobo did you get time to find more details about the problem you are experimenting? Any help here would be useful in driving this one to completion.

Reviewing it again this moment and I found the reason why the PRs fail in my setup... I use the upf-adapter by default and it looks like changes are required for the upf-adapter to handle the new way TEID is allocated. As soon as I disabled the upf-adapter, all my tests were successful. Can you please address this breaking change in the upf-adapter?

Also, I am moving the codebase to GO 1.23. This is because the latest GRPC version (1.68.0) requires any of the latest 2 major GO versions (i.e., 1.22 and/or 1.23). So, I am making this change across all of our repos such that all of them use the same Go version. Can you please make this change in as part of your 2 PRs (UPF and SMF)? or would you like me to open a PR for that and then you rebase your PRs? The key piece would be to have/use GO 1.23 as part of release images v2.0.0

pfcpiface/session_pdr.go

gruyaume · 2024-11-15T18:41:18Z

today

@gab-arrobo did you get time to find more details about the problem you are experimenting? Any help here would be useful in driving this one to completion.

Reviewing it again this moment and I found the reason why the PRs fail in my setup... I use the upf-adapter by default and it looks like changes are required for the upf-adapter to handle the new way TEID is allocated. As soon as I disabled the upf-adapter, all my tests were successful. Can you please address this breaking change in the upf-adapter?

Also, I am moving the codebase to GO 1.23. This is because the latest GRPC version (1.68.0) requires any of the latest 2 major GO versions (i.e., 1.22 and/or 1.23). So, I am making this change across all of our repos such that all of them use the same Go version. Can you please make this change in as part of your 2 PRs (UPF and SMF)? or would you like me to open a PR for that and then you rebase your PRs? The key piece would be to have/use GO 1.23 as part of release images v2.0.0

No problem, I just bumped it in the last commit.

gab-arrobo · 2024-11-15T18:56:39Z

@gruyaume, you need to "redo" the commit Update pfcpiface/session_pdr.go because it is not signed

gab-arrobo · 2024-11-15T19:07:00Z

No problem, I just bumped it in the last commit.

Thanks!

Signed-off-by: Guillaume Belanger <[email protected]>

pfcpiface/upf.go

thakurajayL · 2024-11-18T01:17:35Z

pfcpiface/fteid.go

+}
+
+// Allocate and return an id in range [minValue, maxValue]
+func (idGenerator *FTEIDGenerator) Allocate() (uint32, error) {


Not efficient.. In case we establish few hundred or thousand tunnels then this will become slow. Anything better we can do?

This is using the exact same algorithm as was used in the SMF. While we may want to change it for something efficient, I would decouple this performance improvement from this change here which is focused on moving f-teid allocation from the SMF to the UPF.

Reference:

https://github.com/omec-project/util/blob/main/idgenerator/idgenerator.go#L40

thakurajayL · 2024-11-18T01:24:37Z

Could you please share pcap when you establish 1 call and also PFCP association. Thanks

thakurajayL · 2024-11-18T01:22:17Z

pfcpiface/parse_pdr.go

-		p.tunnelTEIDMask = 0xFFFFFFFF
-		p.tunnelIP4Dst = ip2int(tunnelIPv4Address)
-		p.tunnelIP4DstMask = 0xFFFFFFFF
+	if fteid.HasCh() {


Could you pls help with this. What does this do?

If the SMF wants the UPF to allocate the F-TEID, which is the case in 3GPP release 16, the SMF sets the CHOOSE flag in the Local F-TEID of its PFCP association setup request. Here, the UPF checks if this flag is set. If it is, it sets UPAllocateFteid to True which will result in the UPF allocating an F-TEID.

From the 3GPP spec:

The SMF shall request the UPF to allocate the F-TEID by setting the CHOOSE flag in the Local F-TEID. And if the PDR(s) is created successfully, the UPF shall return the F-TEID(s) it has assigned to the PDR(s) in the PFCP Session Establishment Response or PFCP Session Modification Response.

pfcpiface/parse_pdr.go

Signed-off-by: Guillaume Belanger <[email protected]>

gruyaume · 2024-11-18T14:03:18Z

Could you please share pcap when you establish 1 call and also PFCP association. Thanks

Ok will do.

gruyaume · 2024-11-18T18:12:00Z

Could you please share pcap when you establish 1 call and also PFCP association. Thanks

Ok will do.

@thakurajayL

Here are the captures taken from the SMF and UPF respectively. You likely will want to filter in only pfcp packets.

smf.tar.gz
upf.tar.gz

I had to tar them because github does not accept pcap files.

gruyaume · 2024-11-22T20:04:06Z

@thakurajayL @gab-arrobo can we complete the review here and in the SMF for this feature?

gab-arrobo · 2024-11-22T20:27:03Z

@thakurajayL @gab-arrobo can we complete the review here and in the SMF for this feature?

@thakurajayL, any further comment/question about this PR (and omec-project/smf#334). Everything looks good to me and these PRs are working fine when having the upfadapter in between them.

gruyaume changed the title ~~feat: return F-TEID in PFCP session establishment response~~ feat: add f-teid allocation support Aug 5, 2024

gruyaume changed the title ~~feat: add f-teid allocation support~~ feat: add F-TEID allocation support Aug 5, 2024

gruyaume mentioned this pull request Aug 6, 2024

[Q] 3GPP version and features support of UPF #824

Open

thakurajayL self-assigned this Oct 5, 2024

gruyaume mentioned this pull request Oct 16, 2024

feat: request UPF to allocate F-TEID omec-project/smf#334

Open

gruyaume marked this pull request as ready for review November 12, 2024 11:39

ghislainbourgeois approved these changes Nov 12, 2024

View reviewed changes

gab-arrobo reviewed Nov 12, 2024

View reviewed changes

docs/configuration-guide.md Outdated Show resolved Hide resolved

pfcpiface/fteid.go Outdated Show resolved Hide resolved

dariofaccin approved these changes Nov 13, 2024

View reviewed changes

gruyaume requested a review from gab-arrobo November 13, 2024 13:15

gab-arrobo reviewed Nov 15, 2024

View reviewed changes

pfcpiface/session_pdr.go Outdated Show resolved Hide resolved

feat: add f-teid allocation support

8f17c7e

Signed-off-by: Guillaume Belanger <[email protected]>

gruyaume force-pushed the dev-fteid branch from 1770ce7 to 8f17c7e Compare November 15, 2024 20:00

Merge branch 'main' into dev-fteid

f1de3d4

gab-arrobo reviewed Nov 16, 2024

View reviewed changes

pfcpiface/upf.go Outdated Show resolved Hide resolved

thakurajayL reviewed Nov 18, 2024

View reviewed changes

gruyaume added 2 commits November 18, 2024 08:51

Merge branch 'main' into dev-fteid

33de3cc

chore: uses else if instead of else and then if

12552a0

Signed-off-by: Guillaume Belanger <[email protected]>

gruyaume requested review from thakurajayL and gab-arrobo November 18, 2024 18:12

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat: add F-TEID allocation support #840

feat: add F-TEID allocation support #840

gruyaume commented Aug 5, 2024 •

edited

Loading

omecproject commented Aug 5, 2024

gab-arrobo commented Oct 17, 2024

gab-arrobo commented Oct 17, 2024

gruyaume commented Oct 17, 2024

gab-arrobo left a comment

gruyaume commented Nov 13, 2024

gruyaume commented Nov 13, 2024

gab-arrobo commented Nov 13, 2024

gruyaume commented Nov 15, 2024

gab-arrobo commented Nov 15, 2024

gruyaume commented Nov 15, 2024

gab-arrobo commented Nov 15, 2024

gab-arrobo commented Nov 15, 2024

thakurajayL Nov 18, 2024

gruyaume Nov 18, 2024

thakurajayL commented Nov 18, 2024

thakurajayL Nov 18, 2024

gruyaume Nov 18, 2024

gruyaume commented Nov 18, 2024

gruyaume commented Nov 18, 2024

gruyaume commented Nov 22, 2024

gab-arrobo commented Nov 22, 2024

feat: add F-TEID allocation support #840

Are you sure you want to change the base?

feat: add F-TEID allocation support #840

Conversation

gruyaume commented Aug 5, 2024 • edited Loading

Description

Summary

Changes in the SMF and UPF are intertwined

Reference

3GPP Release 16

omecproject commented Aug 5, 2024

gab-arrobo commented Oct 17, 2024

gab-arrobo commented Oct 17, 2024

gruyaume commented Oct 17, 2024

gab-arrobo left a comment

Choose a reason for hiding this comment

gruyaume commented Nov 13, 2024

gruyaume commented Nov 13, 2024

gab-arrobo commented Nov 13, 2024

gruyaume commented Nov 15, 2024

gab-arrobo commented Nov 15, 2024

gruyaume commented Nov 15, 2024

gab-arrobo commented Nov 15, 2024

gab-arrobo commented Nov 15, 2024

thakurajayL Nov 18, 2024

Choose a reason for hiding this comment

gruyaume Nov 18, 2024

Choose a reason for hiding this comment

thakurajayL commented Nov 18, 2024

thakurajayL Nov 18, 2024

Choose a reason for hiding this comment

gruyaume Nov 18, 2024

Choose a reason for hiding this comment

gruyaume commented Nov 18, 2024

gruyaume commented Nov 18, 2024

gruyaume commented Nov 22, 2024

gab-arrobo commented Nov 22, 2024

gruyaume commented Aug 5, 2024 •

edited

Loading