Re: [PATCH bpf-next v7 5/7] bpf: Add BPF_F_CPU and BPF_F_ALL_CPUS flags support for percpu_cgroup_storage maps

["Leon Hwang" <leon.hwang@linux.dev>]

>> >> @@ -216,7 +222,7 @@ int bpf_percpu_cgroup_storage_update(struct bpf_map *_map, void *key,
>> >>         int cpu, off = 0;
>> >>         u32 size;
>> >>
>> >> -       if (map_flags != BPF_ANY && map_flags != BPF_EXIST)
>> >> +       if ((u32)map_flags & ~(BPF_ANY | BPF_EXIST | BPF_F_CPU | BPF_F_ALL_CPUS))
>> >>                 return -EINVAL;
>> >
>> > shouldn't bpf_map_check_op_flags() be used here to validate cpu number
>> > and BPF_F_CPU and BPF_F_ALL_CPUS exclusivity?..
>> >
>>
>> bpf_map_check_op_flags() has been called in
>> syscall.c::map_update_elem().
>
> ah, I actually tried to double-check that by looking at earlier
> patches, but still missed that. Never mind then.
>

Sorry for the earlier unclear explanation.

Let me restate:

1. Patch #1 introduces bpf_map_check_op_flags().
2. Patch #1 also updates map_update_elem() to call
   bpf_map_check_op_flags().
3. Patch #2 extends bpf_map_check_op_flags() to validate the CPU flags
   and CPU number.

When updating elements of percpu cgroup_storage maps, map_update_elem()
calls bpf_map_check_op_flags() before
bpf_percpu_cgroup_storage_update() being invoked.

So, the CPU flags and CPU number are already validated in
map_update_elem(), and don’t need to be re-checked here.

Thanks,
Leon