kvm: fix security group conntrack NOTRACK optimization for single-direction allow-all rules by weizhouapache · Pull Request #14100 · apache/cloudstack

weizhouapache · 2026-09-09T14:27:40Z

Description

This PR fixes #14099

The raw-table NOTRACK rules for a VM IP in the cs_notrack/cs_notrack6 ipsets apply to that IP as both source and destination, disabling connection tracking for it in both directions at once. The check that populates these ipsets did not look at rule['ruletype'], so an allow-all rule in only one direction was enough to disable conntrack for the VM.

Connection tracking is needed by the restricted direction: the ESTABLISHED,RELATED fallback rules in the VM default chain admit return traffic for connections opened in the other direction. With conntrack disabled they no longer match, so:

restricted ingress + allow-all egress (e.g. egress, all, 0.0.0.0/0): responses to connections initiated by the VM are dropped before reaching the tap interface (VMs with security groups allowing all egress traffic cannot access the Internet in 4.23.0.0 #14099);
allow-all ingress + restricted egress: replies to connections initiated towards the VM are dropped unless the destination is covered by an egress rule.

Now a VM IP is added to the notrack ipset for an IP family only when ingress has an explicit allow-all rule (all, 0.0.0.0/0 or ::/0) and egress is open too, which means either:

an explicit allow-all egress rule, or
no egress rules for that family (the egress chain then ends in an unconditional ACCEPT).

Ingress	Egress	In `cs_notrack` before	In `cs_notrack` after
no rule / restricted	any	only if egress is explicit allow-all (bug)	no
allow all	restricted	yes (bug)	no
allow all	no rule	yes	yes
allow all	allow all	yes	yes

Types of changes

Breaking change (fix or feature that would cause existing functionality to change)
New feature (non-breaking change which adds functionality)
Bug fix (non-breaking change which fixes an issue)
Enhancement (improves an existing feature and functionality)
Cleanup (Code refactoring and cleanup, that may add test cases)
Build/CI
Test (unit or integration test code)

Feature/Enhancement Scale or Bug Severity

Feature/Enhancement Scale

Major
Minor

Bug Severity

Screenshots (if appropriate):

How Has This Been Tested?

How did you try to break this feature and the system with this change?

…ingle-direction allow-all rule The raw-table NOTRACK rules for a VM IP in the cs_notrack/cs_notrack6 ipsets apply to that IP as both source and destination, disabling connection tracking for it in both directions at once. The check that populates these ipsets did not look at rule['ruletype'], so an allow-all rule in only one direction (e.g. egress all to 0.0.0.0/0) was enough to disable conntrack for the VM, breaking return traffic when the other direction (e.g. ingress) is restricted. Only skip conntrack for a family now when both an ingress allow-all and an egress allow-all rule exist for that family, matching the actual bidirectional effect of the NOTRACK rules.

codecov · 2026-09-09T14:49:13Z

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 20.39%. Comparing base (c77e16a) to head (824f92a).
⚠️ Report is 60 commits behind head on main.

Additional details and impacted files

@@             Coverage Diff              @@
##               main   #14100      +/-   ##
============================================
+ Coverage     19.79%   20.39%   +0.60%     
- Complexity    20020    21283    +1263     
============================================
  Files          6371     6481     +110     
  Lines        575954   588040   +12086     
  Branches      70521    72021    +1500     
============================================
+ Hits         114034   119956    +5922     
- Misses       449486   455052    +5566     
- Partials      12434    13032     +598

Flag	Coverage Δ
uitests	`3.70% <ø> (+0.17%)`	⬆️
unittests	`21.70% <ø> (+0.62%)`	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

weizhouapache · 2026-09-09T14:59:47Z

verified the following cases

Ingress	Egress	Actual egress traffic	`ipset list cs_notrack` - before fix	`ipset list cs_notrack` - after fix
no rule (deny)	no rule	allowed (default ACCEPT)	not in set	not in set
no rule (deny)	allow a (restricted)	allowed only for that rule	not in set	not in set
no rule (deny)	allow all (explicit)	allowed	in set (bug)	not in set
allow a (restricted)	no rule	allowed (default ACCEPT)	not in set	not in set
allow a (restricted)	allow a (restricted)	allowed only for that rule	not in set	not in set
allow a (restricted)	allow all (explicit)	allowed	in set (bug)	not in set
allow all (explicit)	no rule	allowed (default ACCEPT)	in set (correct)	in set (correct)
allow all (explicit)	allow a (restricted)	allowed only for that rule	in set (bug)	not in set
allow all (explicit)	allow all (explicit)	allowed	in set (correct)	in set (correct)

Copilot

Warning

Copilot couldn't run its full agentic review because it didn't start before the timeout. Make sure your repository has a runner available, or add a copilot-code-review.yml file specifying one with the runs-on attribute. See the docs for more details.

Pull request overview

Fixes a security group conntrack optimization bug where a single-direction allow-all rule could incorrectly disable conntrack for both traffic directions on a VM IP.

Changes:

Track ingress/egress “allow-all” rules separately per IP family (IPv4/IPv6).
Only disable conntrack (NOTRACK ipset membership) when both ingress and egress are allow-all for a given family.

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

+        ingress4_allow_all = False
+        egress4_allow_all = False
+        ingress6_allow_all = False
+        egress6_allow_all = False


+                if '0.0.0.0/0' in rule['ipv4']:
+                    if rule['ruletype'] == 'E':
+                        egress4_allow_all = True
+                    else:
+                        ingress4_allow_all = True
+                if '::/0' in rule['ipv6']:
+                    if rule['ruletype'] == 'E':
+                        egress6_allow_all = True
+                    else:


nvazquez · 2026-10-05T11:06:28Z

Hi @weizhouapache is this PR ready for review or still in progress?

weizhouapache · 2026-10-05T11:14:11Z

Hi @weizhouapache is this PR ready for review or still in progress?

@nvazquez
it is ready for review

…ss and no egress rules Egress with no rules ends in an unconditional ACCEPT, so it is as open as an explicit allow-all egress rule. Treat it as open when deciding whether the VM IP can be added to the notrack ipset.

weizhouapache · 2026-10-15T14:14:11Z

thanks @Damans227 @vishesh92
makes sense, I updated the PR accordingly

weizhouapache · 2026-10-15T14:15:35Z

@blueorangutan package

blueorangutan · 2026-10-15T14:16:05Z

@weizhouapache a [SL] Jenkins job has been kicked to build packages. It will be bundled with no SystemVM templates. I'll keep you posted as I make progress.

Damans227

thanks @weizhouapache, fix looks right. LGTM

Copilot

Warning

Copilot couldn't run its full agentic review because it didn't start before the timeout. Make sure your repository has a runner available, or add a copilot-code-review.yml file specifying one with the runs-on attribute. See the docs for more details.

Copilot review overview

4 open findings

The PR description says conntrack is only skipped when both an ingress allow-all and an egress… · New
The else branch treats any non-'E' ruletype as ingress. If ruletype ever contains unexpected…
The previous block comment explaining why conntrack could be skipped (and the example rule…
This boolean accumulation pattern is harder to scan than necessary. Consider using idiomatic… · New

🧠 Review effort: Lite

+        # The NOTRACK rules apply to the IP in both directions, so conntrack can only be skipped when
+        # ingress allows all and egress is open as well: either an explicit allow-all egress rule or no
+        # egress rules at all (the egress chain then accepts everything).
+        conntrack4_not_needed = ingress4_allow_all and (egress4_allow_all or not egress4_has_rules)
+        conntrack6_not_needed = ingress6_allow_all and (egress6_allow_all or not egress6_has_rules)


+                egress4_has_rules = egress4_has_rules or bool(rule['ipv4'])
+                egress6_has_rules = egress6_has_rules or bool(rule['ipv6'])


blueorangutan · 2026-10-15T15:16:31Z

Packaging result [SF]: ✔️ el8 ✔️ el9 ✔️ el10 ✔️ debian ✔️ suse15. SL-JID 19567

boring-cyborg Bot added component:networking Python Warning... Python code Ahead! labels Sep 9, 2026

weizhouapache added this to the 24.0 milestone Sep 9, 2026

weizhouapache requested a lite review from Copilot September 9, 2026 15:00

Copilot AI reviewed Sep 9, 2026

View reviewed changes

Copilot started reviewing on behalf of weizhouapache September 9, 2026 20:01

View session

weizhouapache marked this pull request as ready for review October 5, 2026 11:14

Damans227 reviewed Oct 14, 2026

View reviewed changes

Comment thread scripts/vm/network/security_group.py Outdated

Copilot AI lite review requested due to automatic review settings October 15, 2026 14:13

weizhouapache changed the title ~~kvm: fix security group conntrack NOTRACK optimization triggered by single-direction allow-all rule~~ kvm: fix security group conntrack NOTRACK optimization for single-direction allow-all rules Oct 15, 2026

Damans227 approved these changes Oct 15, 2026

View reviewed changes

Copilot AI reviewed Oct 15, 2026

View reviewed changes

Copilot stopped reviewing on behalf of weizhouapache due to an error October 15, 2026 14:59
The job was not acquired by a runner even after multiple attempts.

Navigation Menu

Repository navigation

kvm: fix security group conntrack NOTRACK optimization for single-direction allow-all rules - #14100

kvm: fix security group conntrack NOTRACK optimization for single-direction allow-all rules#14100
weizhouapache wants to merge 2 commits into
apache:mainfrom
weizhouapache:24-fix-sg-allow-all-egress

weizhouapache commented Sep 9, 2026 •

edited

Loading

Uh oh!

codecov Bot commented Sep 9, 2026 •

edited

Loading

Uh oh!

weizhouapache commented Sep 9, 2026 •

edited

Loading

Uh oh!

Copilot AI left a comment

Uh oh!

nvazquez commented Oct 5, 2026 •

edited

Loading

Uh oh!

weizhouapache commented Oct 5, 2026

Uh oh!

Uh oh!

weizhouapache commented Oct 15, 2026

Uh oh!

weizhouapache commented Oct 15, 2026

Uh oh!

blueorangutan commented Oct 15, 2026

Uh oh!

Damans227 left a comment

Uh oh!

Copilot AI left a comment

Uh oh!

blueorangutan commented Oct 15, 2026

Uh oh!

This branch has not been deployed

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

6 participants

		egress4_has_rules = egress4_has_rules or bool(rule['ipv4'])
		egress6_has_rules = egress6_has_rules or bool(rule['ipv6'])

Conversation

weizhouapache commented Sep 9, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Description

Types of changes

Feature/Enhancement Scale or Bug Severity

Feature/Enhancement Scale

Bug Severity

Screenshots (if appropriate):

How Has This Been Tested?

How did you try to break this feature and the system with this change?

Uh oh!

codecov Bot commented Sep 9, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

weizhouapache commented Sep 9, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

nvazquez commented Oct 5, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

weizhouapache commented Oct 5, 2026

Uh oh!

Uh oh!

weizhouapache commented Oct 15, 2026

Uh oh!

weizhouapache commented Oct 15, 2026

Uh oh!

blueorangutan commented Oct 15, 2026

Uh oh!

Damans227 left a comment

Choose a reason for hiding this comment

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Copilot review overview

Uh oh!

blueorangutan commented Oct 15, 2026

Uh oh!

This branch has not been deployed

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

6 participants

weizhouapache commented Sep 9, 2026 •

edited

Loading

codecov Bot commented Sep 9, 2026 •

edited

Loading

weizhouapache commented Sep 9, 2026 •

edited

Loading

nvazquez commented Oct 5, 2026 •

edited

Loading