Apparently, only a small subset of cgroup hooks actually falls
back to cgroup_base_func_proto. This leads to unexpected result
where not all cgroup helpers have bpf_{g,s}et_retval.
It's getting harder and harder to manage which helpers are exported
to which hooks. We now have the following call chains:
- cg_skb_func_proto
- sk_filter_func_proto
- bpf_sk_base_func_proto
- bpf_base_func_proto
So by looking at cg_skb_func_proto it's pretty hard to understand
what's going on.
For cgroup helpers, I'm proposing we do the following instead:
func_proto = cgroup_common_func_proto();
if (func_proto) return func_proto;
/* optional, if hook has 'current' */
func_proto = cgroup_current_func_proto();
if (func_proto) return func_proto;
...
switch (func_id) {
/* hook specific helpers */
case BPF_FUNC_hook_specific_helper:
return &xyz;
default:
/* always fall back to plain bpf_base_func_proto */
bpf_base_func_proto(func_id);
}
If this turns out more workable, we can follow up with converting
the rest to the same pattern.
Stanislav Fomichev (3):
bpf: introduce cgroup_{common,current}_func_proto
bpf: use cgroup_{common,current}_func_proto in more hooks
selftests/bpf: make sure bpf_{g,s}et_retval is exposed everywhere
include/linux/bpf.h | 16 +++
kernel/bpf/cgroup.c | 82 +++++++--------
kernel/bpf/helpers.c | 99 ++++++++++++++++++-
net/core/filter.c | 78 ++++++---------
tools/testing/selftests/bpf/Makefile | 1 +
.../bpf/cgroup_getset_retval_hooks.h | 25 +++++
.../bpf/prog_tests/cgroup_getset_retval.c | 48 +++++++++
.../bpf/progs/cgroup_getset_retval_hooks.c | 16 +++
8 files changed, 270 insertions(+), 95 deletions(-)
create mode 100644 tools/testing/selftests/bpf/cgroup_getset_retval_hooks.h
create mode 100644 tools/testing/selftests/bpf/progs/cgroup_getset_retval_hooks.c
--
2.37.1.595.g718a3a8f04-goog
