Re: [PATCH bpf-next v3 4/6] mm: introduce BPF kfuncs to access memcg statistics and events

[Roman Gushchin]

Alexei Starovoitov <alexei.starovoitov@xxxxxxxxx> writes:

> On Mon, Dec 22, 2025 at 12:18 PM Roman Gushchin
> <roman.gushchin@xxxxxxxxx> wrote:
>>
>> +       if (idx < 0 || idx >= MEMCG_NR_STAT || !memcg_stat_item_valid(idx))
>> +               return (unsigned long)-1;
>
> memcg_stat_item_valid() and memcg_stat_item_valid()
> helpers introduced specifically to be used in these kfuncs,
> so I feel it's cleaner to do all idx checking within them
> instead of splitting the checks like this.
> Then it will be easier to see that
> memcg_stat_item_valid(idx) access is in bounds when idx < MEMCG_NR_STAT
>
> Also I'd do one check like (u32)idx >= MEMCG_NR_STAT
> and drop idx < 0 part. Compiler is probably smart enough to
> optimize this way itself, but I'd still do one check.

Sure, good point.
Thanks for reviews!