Ihor Solodrai reported selftest 'btf_tag/btf_type_tag_percpu_vmlinux_helper'
failure ([1]) during 6.16 merge window. The failure log:
...
7: (15) if r0 == 0x0 goto pc+1 ; R0=ptr_css_rstat_cpu()
; *(volatile int *)rstat; @ btf_type_tag_percpu.c:68
8: (61) r1 = *(u32 *)(r0 +0)
cannot access ptr member updated_children with moff 0 in struct css_rstat_cpu with off 0 size 4
Two changes are needed. First, 'struct cgroup_rstat_cpu' needs to be
replaced with 'struct css_rstat_cpu' to be consistent with new data
structure. Second, layout of 'css_rstat_cpu' is changed compared
to 'cgroup_rstat_cpu'. The first member becomes a pointer so
the bpf prog needs to do 8-byte load instead of 4-byte load.
[1] https://lore.kernel.org/bpf/
6f688f2e-7d26-423a-9029-
d1b1ef1c938a@linux.dev/
Cc: Ihor Solodrai <ihor.solodrai@linux.dev>
Cc: JP Kobryn <inwardvessel@gmail.com>
Signed-off-by: Yonghong Song <yonghong.song@linux.dev>
Acked-by: JP Kobryn <inwardvessel@gmail.com>
Link: https://lore.kernel.org/r/20250529201151.1787575-1-yonghong.song@linux.dev
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
SEC("tp_btf/cgroup_mkdir")
int BPF_PROG(test_percpu_helper, struct cgroup *cgrp, const char *path)
{
- struct cgroup_rstat_cpu *rstat;
+ struct css_rstat_cpu *rstat;
__u32 cpu;
cpu = bpf_get_smp_processor_id();
- rstat = (struct cgroup_rstat_cpu *)bpf_per_cpu_ptr(
+ rstat = (struct css_rstat_cpu *)bpf_per_cpu_ptr(
cgrp->self.rstat_cpu, cpu);
if (rstat) {
/* READ_ONCE */
- *(volatile int *)rstat;
+ *(volatile long *)rstat;
}
return 0;