On 6/27/25 2:38 PM, Alexei Starovoitov wrote:
On Fri, Jun 27, 2025 at 2:36 PM Ihor Solodrai <ihor.solodrai@xxxxxxxxx> wrote:
On 6/27/25 2:34 PM, Alexei Starovoitov wrote:
On Fri, Jun 27, 2025 at 2:19 PM Ihor Solodrai <ihor.solodrai@xxxxxxxxx> wrote:
On 6/27/25 12:12 PM, Song Liu wrote:
cgroup_xattr/read_cgroupfs_xattr has two issues:
1. cgroup_xattr/read_cgroupfs_xattr messes up lo without creating a netns
first. This causes issue with other tests.
Fix this by using a different hook (lsm.s/file_open) and not messing
with lo.
2. cgroup_xattr/read_cgroupfs_xattr sets up cgroups without proper
mount namespaces.
Fix this by using the existing cgroup helpers. A new helper
set_cgroup_xattr() is added to set xattr on cgroup files.
Fixes: f4fba2d6d282 ("selftests/bpf: Add tests for bpf_cgroup_read_xattr")
Reported-by: Alexei Starovoitov <ast@xxxxxxxxxx>
Closes: https://lore.kernel.org/bpf/CAADnVQ+iqMi2HEj_iH7hsx+XJAsqaMWqSDe4tzcGAnehFWA9Sw@xxxxxxxxxxxxxx/
Signed-off-by: Song Liu <song@xxxxxxxxxx>
---
Changes v1 => v2:
1. Add the second fix above.
v1: https://lore.kernel.org/bpf/20250627165831.2979022-1-song@xxxxxxxxxx/
---
tools/testing/selftests/bpf/cgroup_helpers.c | 21 ++++
tools/testing/selftests/bpf/cgroup_helpers.h | 4 +
.../selftests/bpf/prog_tests/cgroup_xattr.c | 117 ++++--------------
.../selftests/bpf/progs/read_cgroupfs_xattr.c | 4 +-
4 files changed, 49 insertions(+), 97 deletions(-)
Hi Song.
I tried this patch on BPF CI, and it appears it fixes the hanging
failure we've been seeing today on bpf-next and netdev.
I am going to add it to ci/diffs.
Applied to bpf-next already.
CI patches apply to all base branches. My understanding is, it's needed
at least for netdev too.
How is that possible?
The offending commit is only in /master and in /for-next branches,
while /for-next is there for linux-next only.
Alexei, for-next contains offending commit, but does not have Song's
fix. Right now it's the only base branch on BPF CI that uses the temp
patch.
We do run tests on for-next, so I suppose the patch should remain in
ci/diffs until it's committed into for-next?
