On Thu, Feb 9, 2017 at 4:28 AM, Tariq Toukan <ttoukan.li...@gmail.com> wrote:
>
>
> On 07/02/2017 5:02 AM, Eric Dumazet wrote:
>>
>> + for (nr = 0;;) {
>> + frag_size = min_t(int, length, frag_info->frag_size);
>> +
>> if (unlikely(!frags[nr].page))
>> goto fail;
>> @@ -603,15 +602,16 @@ static int mlx4_en_complete_rx_desc(struct
>> mlx4_en_priv *priv,
>> __skb_fill_page_desc(skb, nr, frags[nr].page,
>> frags[nr].page_offset,
>> - frag_info->frag_size);
>> + frag_size);
>
> Same for dma_sync_single (two lines above).
Oh right, although I was not sure it this really matter on a part of a page.
