Re: [PATCH v2 bpf-next 1/3] bpf: fix BTF verifier size resolution logic

[Yonghong Song <yhs@fb.com>]

On 7/12/19 8:36 AM, Andrii Nakryiko wrote:
> On Thu, Jul 11, 2019 at 10:59 PM Yonghong Song <yhs@fb.com> wrote:
>>
>>
>>
>> On 7/10/19 11:53 PM, Andrii Nakryiko wrote:
>>> BTF verifier has a size resolution bug which in some circumstances leads to
>>> invalid size resolution for, e.g., TYPEDEF modifier.  This happens if we have
>>> [1] PTR -> [2] TYPEDEF -> [3] ARRAY, in which case due to being in pointer
>>> context ARRAY size won't be resolved (because for pointer it doesn't matter, so
>>> it's a sink in pointer context), but it will be permanently remembered as zero
>>> for TYPEDEF and TYPEDEF will be marked as RESOLVED. Eventually ARRAY size will
>>> be resolved correctly, but TYPEDEF resolved_size won't be updated anymore.
>>> This, subsequently, will lead to erroneous map creation failure, if that
>>> TYPEDEF is specified as either key or value, as key_size/value_size won't
>>> correspond to resolved size of TYPEDEF (kernel will believe it's zero).
>>>
>>> Note, that if BTF was ordered as [1] ARRAY <- [2] TYPEDEF <- [3] PTR, this
>>> won't be a problem, as by the time we get to TYPEDEF, ARRAY's size is already
>>> calculated and stored.
>>>
>>> This bug manifests itself in rejecting BTF-defined maps that use array
>>> typedef as a value type:
>>>
>>> typedef int array_t[16];
>>>
>>> struct {
>>>       __uint(type, BPF_MAP_TYPE_ARRAY);
>>>       __type(value, array_t); /* i.e., array_t *value; */
>>> } test_map SEC(".maps");
>>>
>>> The fix consists on not relying on modifier's resolved_size and instead using
>>> modifier's resolved_id (type ID for "concrete" type to which modifier
>>> eventually resolves) and doing size determination for that resolved type. This
>>> allow to preserve existing "early DFS termination" logic for PTR or
>>> STRUCT_OR_ARRAY contexts, but still do correct size determination for modifier
>>> types.
>>>
>>> Fixes: eb3f595dab40 ("bpf: btf: Validate type reference")
>>> Cc: Martin KaFai Lau <kafai@fb.com>
>>> Signed-off-by: Andrii Nakryiko <andriin@fb.com>
>>> ---
>>>    kernel/bpf/btf.c | 14 ++++++++++----
>>>    1 file changed, 10 insertions(+), 4 deletions(-)
>>>
>>> diff --git a/kernel/bpf/btf.c b/kernel/bpf/btf.c
>>> index cad09858a5f2..22fe8b155e51 100644
>>> --- a/kernel/bpf/btf.c
>>> +++ b/kernel/bpf/btf.c
>>> @@ -1073,11 +1073,18 @@ const struct btf_type *btf_type_id_size(const struct btf *btf,
>>>                                 !btf_type_is_var(size_type)))
>>>                        return NULL;
>>>
>>> -             size = btf->resolved_sizes[size_type_id];
>>>                size_type_id = btf->resolved_ids[size_type_id];
>>>                size_type = btf_type_by_id(btf, size_type_id);
>>>                if (btf_type_nosize_or_null(size_type))
>>>                        return NULL;
>>> +             else if (btf_type_has_size(size_type))
>>> +                     size = size_type->size;
>>> +             else if (btf_type_is_array(size_type))
>>> +                     size = btf->resolved_sizes[size_type_id];
>>> +             else if (btf_type_is_ptr(size_type))
>>> +                     size = sizeof(void *);
>>> +             else
>>> +                     return NULL;
>>
>> Looks good to me. Not sure whether we need to do any adjustment for
>> var kind or not. Maybe we can do similar change in btf_var_resolve()
> 
> I don't think btf_var_resolve() needs any change. btf_var_resolve
> can't be referenced by modifier, so it doesn't have any problem. It's
> similar to array in that it's size will be determined correctly.

Correct. With your previous patch, the resolved_sizes[..] for var type 
is not used, so that is why I suggest to just set it to 0.

> 
> But I think btf_type_id_size() doesn't handle var case correctly, I'll do
> 
> +             else if (btf_type_is_array(size_type) ||
> btf_type_is_var(size_type))
> +                     size = btf->resolved_sizes[size_type_id];

This change should work too (to use btf->resolved_sizes[...]).

> 
> to fix that.
> 
>> to btf_modifier_resolve()? But I do not think it impacts correctness
>> similar to btf_modifier_resolve() below as you changed
>> btf_type_id_size() implementation in the above.
>>
>>>        }
>>>
>>>        *type_id = size_type_id;
>>> @@ -1602,7 +1609,6 @@ static int btf_modifier_resolve(struct btf_verifier_env *env,
>>>        const struct btf_type *next_type;
>>>        u32 next_type_id = t->type;
>>>        struct btf *btf = env->btf;
>>> -     u32 next_type_size = 0;
>>>
>>>        next_type = btf_type_by_id(btf, next_type_id);
>>>        if (!next_type || btf_type_is_resolve_source_only(next_type)) {
>>> @@ -1620,7 +1626,7 @@ static int btf_modifier_resolve(struct btf_verifier_env *env,
>>>         * save us a few type-following when we use it later (e.g. in
>>>         * pretty print).
>>>         */
>>> -     if (!btf_type_id_size(btf, &next_type_id, &next_type_size)) {
>>> +     if (!btf_type_id_size(btf, &next_type_id, NULL)) {
>>>                if (env_type_is_resolved(env, next_type_id))
>>>                        next_type = btf_type_id_resolve(btf, &next_type_id);
>>>
>>> @@ -1633,7 +1639,7 @@ static int btf_modifier_resolve(struct btf_verifier_env *env,
>>>                }
>>>        }
>>>
>>> -     env_stack_pop_resolved(env, next_type_id, next_type_size);
>>> +     env_stack_pop_resolved(env, next_type_id, 0);
>>>
>>>        return 0;
>>>    }
>>>