Re: [PATCH net-next 1/3] net: ipv6: ioam6_iptunnel: mitigate 2-realloc issue

[Alexander Lobakin]

From: Justin Iurman <justin.iurman@xxxxxxxxx>
Date: Fri, 25 Oct 2024 23:06:36 +0200

> On 10/25/24 17:12, Alexander Lobakin wrote:
>> From: Justin Iurman <justin.iurman@xxxxxxxxx>
>> Date: Fri, 25 Oct 2024 15:37:25 +0200
>>
>>> This patch mitigates the two-reallocations issue with ioam6_iptunnel by
>>> providing the dst_entry (in the cache) to the first call to
>>> skb_cow_head(). As a result, the very first iteration would still
>>> trigger two reallocations (i.e., empty cache), while next iterations
>>> would only trigger a single reallocation.
>>
>> [...]
>>
>>>   static int ioam6_do_inline(struct net *net, struct sk_buff *skb,
>>> -               struct ioam6_lwt_encap *tuninfo)
>>> +               struct ioam6_lwt_encap *tuninfo,
>>> +               struct dst_entry *dst)
>>>   {
>>>       struct ipv6hdr *oldhdr, *hdr;
>>>       int hdrlen, err;
>>>         hdrlen = (tuninfo->eh.hdrlen + 1) << 3;
>>>   -    err = skb_cow_head(skb, hdrlen + skb->mac_len);
>>> +    err = skb_cow_head(skb, hdrlen + (!dst ? skb->mac_len
>>> +                           : LL_RESERVED_SPACE(dst->dev)));
>>
>> You use this pattern a lot throughout the series. I believe you should
>> make a static inline or a macro from it.
>>
>> static inline u32 some_name(const *dst, const *skb)
>> {
>>     return dst ? LL_RESERVED_SPACE(dst->dev) : skb->mac_len;
>> }
>>
>> BTW why do you check for `!dst`, not `dst`? Does changing this affects
>> performance?
> 
> Not at all, you're right... even the opposite actually. Regarding the
> static inline suggestion, it could be a good idea and may even look like
> this as an optimization:
> 
> static inline u32 dev_overhead(struct dst_entry *dst, struct sk_buff *skb)
> {
>     if (likely(dst))
>         return LL_RESERVED_SPACE(dst->dev);
> 
>     return skb->mac_len;
> }

Oh, nice!

> 
> The question is... where should it go then? A static inline function per
> file (i.e., ioam6_iptunnel.c, seg6_iptunnel.c, and rpl_iptunnel.c)? In
> that case, it would still be repeated 3 times. Or in a header file
> somewhere, to have it defined only once? If so, what location do you
> think would be best?

100% should be in a header file. Can't suggest any since I don't usually
work with tunnels and ain't deep into its header structure.

Thanks,
Olek