Bug 2447287
| Summary: | Unpacking sources end with "/usr/bin/tar: *: Cannot open: Function not implemented" | ||
|---|---|---|---|
| Product: | [Fedora] Fedora | Reporter: | Marcin Juszkiewicz <mjuszkie> |
| Component: | tar | Assignee: | Pavel Raiskup <praiskup> |
| Status: | CLOSED DUPLICATE | QA Contact: | Fedora Extras Quality Assurance <extras-qa> |
| Severity: | unspecified | Docs Contact: | |
| Priority: | unspecified | ||
| Version: | 44 | CC: | copr-team, jamartis, jnovy, kdudka, ljavorsk, msuchy, ovasik, pcahyna, pkubat, praiskup, rjones, williams |
| Target Milestone: | --- | Keywords: | Regression, Upgrades |
| Target Release: | --- | ||
| Hardware: | aarch64 | ||
| OS: | Linux | ||
| Whiteboard: | |||
| Fixed In Version: | Doc Type: | --- | |
| Doc Text: | Story Points: | --- | |
| Clone Of: | Environment: | ||
| Last Closed: | 2026-03-13 15:51:52 UTC | Type: | --- |
| Regression: | --- | Mount Type: | --- |
| Documentation: | --- | CRM: | |
| Verified Versions: | Category: | --- | |
| oVirt Team: | --- | RHEL 7.3 requirements from Atomic Host: | |
| Cloudforms Team: | --- | Target Upstream Version: | |
| Embargoed: | |||
|
Description
Marcin Juszkiewicz
2026-03-13 10:19:26 UTC
Happens with any package, not only kernel. And for kernel I can workaround it by building it for Fedora 43. Mock is using a tooling from the target distribution; namely rpmbuild.fc44 that is using xz.f44 (where I guess happened some change). Function not implemented (ENOSYS) may indicate that kernel system calls are being filtered. Is seccomp involved? Cgroups / Containers? The only recent change to xz I'm aware of affects only containers being run on a RHEL 9 kernel (which does not sound like it's involved here). Also it has a very distinctive error message, also not seen here. For reference that would be: https://bugzilla.redhat.com/show_bug.cgi?id=2407105 Also can you tell me the exact xz NVR where it was working vs failed, also the exact kernel version, and any other relevant information about sandboxing, especially things that might have changed between F43 and F44. root.log of fc44 failed builds says: DEBUG util.py:463: xz-libs-5.8.2-2.fc44.aarch64 DEBUG util.py:463: xz-5.8.2-2.fc44.aarch64 DEBUG util.py:463: xz-devel-5.8.2-2.fc44.aarch64 working fc43 build (on same f44 host) says: DEBUG util.py:463: xz-libs-5.8.1-4.fc43.aarch64 DEBUG util.py:463: xz-5.8.1-4.fc43.aarch64 DEBUG util.py:463: xz-devel-5.8.1-4.fc43.aarch64 Kernel is self-built 6.19.7-300.fc43.pcie65.6.aarch64 (6.19.7 from Fedora kernel/f44 branch + two Ampere PCIE-65 patches, built in fedora-43-aarch64 mock). tar-1.35-8.fc44.aarch64 vs tar-1.35-6.fc43.aarch64 I couldn't reproduce this on Fedora 43 / aarch64. I'm updating my instance to Fedora 44 now. What is the version of mock where this fails? This sure looks suspicious: https://src.fedoraproject.org/rpms/tar/c/5a186e814f50ab1c905d0d14273abc126e2e9509?branch=rawhide $ rpm -qa|grep ^mock mock-filesystem-6.7-1.fc44.noarch mock-6.7-1.fc44.noarch mock-core-configs-44.2-1.fc44.noarch Hi @mjuszkie are you using nspawn option --suppress-sync=yes ? If so, try removing it from Mock config. Please see bz2437037, in particular https://bugzilla.redhat.com/show_bug.cgi?id=2437037#c3 . So we believe this works with tar on Fedora 43, and fails starting with tar on Fedora 44. tar on Fedora 44 contains the openat2 patch (see above). However it might still be something else that is filtering out the openat2 system call, rather than a problem with tar itself. Another thought: is your AArch64 virtualized? There was a related bug in Qemu : https://gitlab.com/qemu-project/qemu/-/work_items/3262 *** This bug has been marked as a duplicate of bug 2437037 *** Thanks Richard and Pavel for help - it was mock option. Pavel: no, it is bare-metal host. Thank you Marcin for the quick confirmation! |