On Sun, May 24, 2020 at 9:09 PM Jan Stancek <[email protected]> wrote:
> ----- Original Message -----
> > > FYI: This (and similar s390x ARK selinux-testuite failures) is caused
> > > by a BPF/BTF-related issue, presumably caused/exposed by the CI ARK
> > > kernels being cross-compiled. It is being investigated.
> >
> > Thanks for the note, the LTP failures are also BPF related and started only
> > recently as well, while BPF testcases haven't changed since March.
>
> It does look like endian issue, presumably related to pahole (CC Arnaldo)
> Only failing arch is s390, which is big endian.
>
> struct btf_header {
>     __u16   magic;
>     __u8    version;
>     __u8    flags;
>     __u32   hdr_len;
> ...
>
> CKI cross-compiled kernel:
>
> # uname -r
> 5.7.0-0.rc6.1.cki.fc33.s390x
> # hexdump -C /sys/kernel/btf/vmlinux | head -1
> 00000000  9f eb 01 00 18 00 00 00  00 00 00 00 a4 9f 18 00  |................|
>
> Build from koji:
>
> # uname -r
> 5.7.0-0.rc6.20200522git051143e1602d.1.fc33.s390x
> # hexdump -C /sys/kernel/btf/vmlinux | head -1
> 00000000  eb 9f 01 00 00 00 00 18  00 00 00 00 00 18 35 ac  |..............5.|
>
> bpf(BPF_PROG_LOAD, {prog_type=BPF_PROG_TYPE_SOCKET_FILTER, insn_cnt=1, 
> insns=[{code=BPF_JMP|BPF_K|BPF_EXIT, dst_reg=BPF_REG_0, src_reg=BPF_REG_0, 
> off=0, imm=0}], license="GP
> L", log_level=7, log_size=8192, log_buf="", kern_version=KERNEL_VERSION(0, 0, 
> 0), prog_flags=0, prog_name="", prog_ifindex=0, 
> expected_attach_type=BPF_CGROUP_INET_INGRESS, prog_btf_fd=0, f
> unc_info_rec_size=0, func_info=NULL, func_info_cnt=0, line_info_rec_size=0, 
> line_info=NULL, line_info_cnt=0, attach_btf_id=0}, 112) = -1 EINVAL (Invalid 
> argument)
>
> Failed verification: in-kernel BTF is malformed

One other thing that might be related to this: The BPF test then
started to fail also on 5.6 CKI stable kernels now (not at the same
time as ARK) and it coincides with CONFIG_DEBUG_INFO=y starting to
appear in the s390x configs (it is the only s390x config option that
changed between the last passing and first failing stable CKI build).

-- 
Ondrej Mosnacek <omosnace at redhat dot com>
Software Engineer, Security Technologies
Red Hat, Inc.
_______________________________________________
kernel mailing list -- [email protected]
To unsubscribe send an email to [email protected]
Fedora Code of Conduct: 
https://docs.fedoraproject.org/en-US/project/code-of-conduct/
List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines
List Archives: 
https://lists.fedoraproject.org/archives/list/[email protected]

Reply via email to