Re: [PATCH bpf 2/9] bpf: Acquire map uref in .init_seq_private for hash map iterator

[Yonghong Song <yhs@fb.com>]

On 8/6/22 12:40 AM, Hou Tao wrote:
> From: Hou Tao <houtao1@huawei.com>
> 
> During bpf(BPF_LINK_CREATE) for BPF_TRACE_ITER, bpf_iter_attach_map()
> has already acquired a map uref, but the uref may be released
> by bpf_link_release() during th reading of map iterator.
> 
> So acquiring an extra map uref in bpf_iter_init_hash_map() and
> releasing it in bpf_iter_fini_hash_map().
> 
> Fixes: d6c4503cc296 ("bpf: Implement bpf iterator for hash maps")
> Signed-off-by: Hou Tao <houtao1@huawei.com>

See my previous reply for some wording issue.

Acked-by: Yonghong Song <yhs@fb.com>

> ---
>   kernel/bpf/hashtab.c | 2 ++
>   1 file changed, 2 insertions(+)
> 
> diff --git a/kernel/bpf/hashtab.c b/kernel/bpf/hashtab.c
> index da7578426a46..da8c0177f773 100644
> --- a/kernel/bpf/hashtab.c
> +++ b/kernel/bpf/hashtab.c
> @@ -2064,6 +2064,7 @@ static int bpf_iter_init_hash_map(void *priv_data,
>   		seq_info->percpu_value_buf = value_buf;
>   	}
>   
> +	bpf_map_inc_with_uref(map);
>   	seq_info->map = map;
>   	seq_info->htab = container_of(map, struct bpf_htab, map);
>   	return 0;
> @@ -2073,6 +2074,7 @@ static void bpf_iter_fini_hash_map(void *priv_data)
>   {
>   	struct bpf_iter_seq_hash_map_info *seq_info = priv_data;
>   
> +	bpf_map_put_with_uref(seq_info->map);
>   	kfree(seq_info->percpu_value_buf);
>   }
>