Micro support

[marcosps]

No description provided.

[fgyanz]

Hi, would it be possible to also include a test? As Sle Micro is kind of an exception among all the codestreams.

[marcosps]

Absolutely. I forgot to do it, but I'll prepare something in the next few days.

[marcosps]

@fgyanz @vmezzela new version! Would you like to add this one first, before the code reorganization?

[vmezzela]

Sure, I can rebase #62 on top of this.

[marcosps]

Rebased on top of current devel (after merging #63)

[vmezzela]

FWIW, I believe this breaks when loading codestream.json files that were stored prior to this patch. I don't know if this can be an issue.

[marcosps]

Yes, I should have mentioned. We already had changes like this in the past, so I always suggest that people first finish their livepatch, and then apply the changes when starting to work on something new. I trully hope this is the last of the breaking changes that we need for MICRO...

[vmezzela]

No problems on my side then!

[fgyanz]

Hi @marcosps ,
PR tested and it's looking pretty good. I think I've only found one issue.

• After running:

lp-build setup --cve 2024-36971 -n bsc1226324 --conf CONFIG_IPV6 --file-funcs net/ipv6/route.c ip6_negative_advice --file-funcs net/ipv4/route.c ipv4_negative_advice --file-funcs net/xfrm/xfrm_policy.c xfrm_negative_advice --file-funcs net/core/sock.c sock_setsockopt sk_setsockopt --file-funcs net/ipv4/tcp_timer.c tcp_retransmit_timer --no-check

I noticed that the generated LP files externalize the symbols using the old method:
In klp/livepatches/bsc1226324/ccp/6.0u2/lp/bsc1226324_net_core_sock.c

static struct klp_kallsyms_reloc klp_funcs[] = {
»       { "mem_cgroup_charge_skmem", (void *)&klpe_mem_cgroup_charge_skmem },
»       { "mem_cgroup_uncharge_skmem",
»         (void *)&klpe_mem_cgroup_uncharge_skmem },
»       { "reuseport_update_incoming_cpu",
»         (void *)&klpe_reuseport_update_incoming_cpu },
»       { "sk_attach_bpf", (void *)&klpe_sk_attach_bpf },
»       { "sk_reuseport_attach_bpf", (void *)&klpe_sk_reuseport_attach_bpf },
»       { "sk_reuseport_attach_filter",
»         (void *)&klpe_sk_reuseport_attach_filter },
»       { "sock_bindtoindex_locked", (void *)&klpe_sock_bindtoindex_locked },
»       { "sock_set_timeout", (void *)&klpe_sock_set_timeout },
»       { "sock_set_timestamp", (void *)&klpe_sock_set_timestamp },
»       { "sock_set_timestamping", (void *)&klpe_sock_set_timestamping },
};

As a side note, and just for the record, I tried running scan with a new CVE, which I personally made sure it affects one Micro codestream and not the following update. The results are as expected :)

klp-build scan --cve 2024-46772
...

Skipping already patched codestreams:
	6.0rtu3 6.0u3 15.6rtu4-6 15.6u5-6
Skipping unaffected codestreams (missing backports):
	12.5u52-63 15.2u44-55 15.3u40-51 15.4u22-34 15.5u9-21
Checking filter and skips...
All affected codestreams:
	6.0rtu2 6.0u2 15.6rtu0-3 15.6u0-4

[marcosps]

Hi @marcosps , PR tested and it's looking pretty good. I think I've only found one issue.

• After running:

lp-build setup --cve 2024-36971 -n bsc1226324 --conf CONFIG_IPV6 --file-funcs net/ipv6/route.c ip6_negative_advice --file-funcs net/ipv4/route.c ipv4_negative_advice --file-funcs net/xfrm/xfrm_policy.c xfrm_negative_advice --file-funcs net/core/sock.c sock_setsockopt sk_setsockopt --file-funcs net/ipv4/tcp_timer.c tcp_retransmit_timer --no-check

I noticed that the generated LP files externalize the symbols using the old method: In klp/livepatches/bsc1226324/ccp/6.0u2/lp/bsc1226324_net_core_sock.c

static struct klp_kallsyms_reloc klp_funcs[] = {
»       { "mem_cgroup_charge_skmem", (void *)&klpe_mem_cgroup_charge_skmem },
»       { "mem_cgroup_uncharge_skmem",
»         (void *)&klpe_mem_cgroup_uncharge_skmem },
»       { "reuseport_update_incoming_cpu",
»         (void *)&klpe_reuseport_update_incoming_cpu },
»       { "sk_attach_bpf", (void *)&klpe_sk_attach_bpf },
»       { "sk_reuseport_attach_bpf", (void *)&klpe_sk_reuseport_attach_bpf },
»       { "sk_reuseport_attach_filter",
»         (void *)&klpe_sk_reuseport_attach_filter },
»       { "sock_bindtoindex_locked", (void *)&klpe_sock_bindtoindex_locked },
»       { "sock_set_timeout", (void *)&klpe_sock_set_timeout },
»       { "sock_set_timestamp", (void *)&klpe_sock_set_timestamp },
»       { "sock_set_timestamping", (void *)&klpe_sock_set_timestamping },
};

As a side note, and just for the record, I tried running scan with a new CVE, which I personally made sure it affects one Micro codestream and not the following update. The results are as expected :)

klp-build scan --cve 2024-46772 ...

Skipping already patched codestreams:
	6.0rtu3 6.0u3 15.6rtu4-6 15.6u5-6
Skipping unaffected codestreams (missing backports):
	12.5u52-63 15.2u44-55 15.3u40-51 15.4u22-34 15.5u9-21
Checking filter and skips...
All affected codestreams:
	6.0rtu2 6.0u2 15.6rtu0-3 15.6u0-4

(@fgyanz Instead of quoting your comment I edited it, sorry....)

Ouch... thanks a lot for catching this mistake! MICRO also uses IBT, and I forgot to tag as such. I've fixed it in the latest push.

[fgyanz]

LGTM! Code reviewed and tested.