Added utility ipv6_is_mc_mac()

@gentoo-root on the complexity issue, pushed a revised version + , the diff from the original commit:

~/dev/cilium2/bpf/tests$ git diff HEAD^ -- ../lib/
diff --git a/bpf/lib/ipv6.h b/bpf/lib/ipv6.h
index c358ddedea..1e25aa82ac 100644
--- a/bpf/lib/ipv6.h
+++ b/bpf/lib/ipv6.h
@@ -197,17 +197,14 @@ static __always_inline bool ipv6_addr_equals(const union v6addr *a,
 static __always_inline
 void ipv6_mc_mac_set(const union v6addr *addr, union macaddr *mac)
 {
-       const union macaddr mac_base_addr = {{0x33, 0x33, 0x00,
-                                             0x00, 0x00, 0x00}};
-
-       memcpy(mac, (void *)&mac_base_addr, 2);
+       mac->addr[0] = mac->addr[1] = 0x33;
        memcpy((__u8 *)mac + 2, (__u8 *)addr + 12, 4);
 }
 
 static __always_inline
 bool ipv6_is_mc_mac(const union v6addr *addr, const union macaddr *mac)
 {
-       union macaddr mc_mac;
+       union macaddr mc_mac __align_stack_8;
 
        ipv6_mc_mac_set(addr, &mc_mac);
        return eth_addrcmp((const union macaddr *)&mc_mac, mac) == 0;
@@ -218,8 +215,7 @@ void ipv6_mc_addr_set(const union v6addr *addr, union v6addr *mc_addr)
 {
        const union v6addr base_addr = { .addr = {0xff, 0x02, 0, 0, 0, 0, 0, 0,
                                          0, 0, 0, 0x01, 0xFF, 0, 0, 0} };
-
-       memcpy(mc_addr, (void *)&base_addr, IPV6_ALEN);
+       *mc_addr = base_addr;
        mc_addr->addr[13] = addr->addr[13];
        mc_addr->addr[14] = addr->addr[14];
        mc_addr->addr[15] = addr->addr[15];

I think this looks much better now. Asking for a re-review, JIC.

Can someone please merge? No more reviews needed, no need for a release note, no need for tests (will only be exercised by #39648.

Thx

Cc @joestringer

/test

no need for tests (will only be exercised by #39648).

Not strictly true. Any refactor can cause build breakage, complexity issues, etc. The rule is: We run the tests. If they're failing, we can investigate. I'll share an example of the triage that can help to identify issues and communicate the status of CI for this specific PR.

Failure triage:

• ci-e2e-upgrade: Filed CI: no-policies-extra/pod-to-local-nodeport: command terminated with exit code 28 #39793 to follow up. It does not appear to be related to the PR.
• ci-integration: failure appears related to Ubuntu APT repository infrastructure availability. Most likely this should recover by itself, not sure we can take any action to resolve.

I retriggered the tests and marked the PR for auto-merge.

Not strictly true. Any refactor can cause build breakage, complexity issues, etc

This is not a refactor. The 3 new functions are not called by any program other than the unit test ipv6_test.c, so they will not even be part of the binary. I believe some smoke tests that run on every PR, so not with /test, build with ipv6 enabled, so I don't see how this can be an issue in this very particular case. I understand and furiously agree in general.

I stand corrected, I scanned the PR too quickly to notice.

Hitting some issues with merge queue; latest is a sigstore infrastructure unavailability: https://github.com/cilium/cilium/actions/runs/15332419707/job/43142145733#step:23:32 . I'll retry.