muvarov replied on github web page:
platform/linux-generic/odp_packet.c
line 18
@@ -1982,12 +1983,31 @@ static inline void parse_tcp(packet_parser_t *prs,
*parseptr += (uint32_t)tcp->hl * 4;
}
+static inline void parse_vxlan(packet_parser_t *prs, packet_parser_t *in_prs,
+ const uint8_t **parseptr, uint32_t *offset,
+ uint32_t frame_len, uint32_t seg_len)
+{
+ const _odp_vxlanhdr_t *vxlan = (const _odp_vxlanhdr_t *)*parseptr;
+
+ if (ODPH_VXLAN_BIT(vxlan->flags))
Comment:
do not use helper in platform code. Theoretically that should not even compile.
(I hope I will add automatic check for that some day.)
> muvarov wrote
> `_ODPH_VXLAN_`
>> muvarov wrote
>> Please make name more common. ODPH_VXLAN_something everywhere.
>>> Balasubramanian Manoharan(bala-manoharan) wrote:
>>> Even if we add the capability currently we don't have ODP APIs expose them
>>> as packet metadata functions. i.e odp_packet_has_inner_tcp()
>>>
>>> Should we add the capability and defer this to after tiger-moth?
>>>> Balasubramanian Manoharan(bala-manoharan) wrote:
>>>> Will migrate to api-next
>>>>> Balasubramanian Manoharan(bala-manoharan) wrote:
>>>>> Okay
>>>>>> Balasubramanian Manoharan(bala-manoharan) wrote:
>>>>>> Will change the typo.
>>>>>>> Bill Fischofer(Bill-Fischofer-Linaro) wrote:
>>>>>>> The "TODO" can't stand by itself. Sounds like we need an additional
>>>>>>> `odp_classification_capability()` bit to indicate whether inner packets
>>>>>>> are included in the parse?
>>>>>>>> Bill Fischofer(Bill-Fischofer-Linaro) wrote:
>>>>>>>> Typo: modified
>>>>>>>>> Bill Fischofer(Bill-Fischofer-Linaro) wrote:
>>>>>>>>> Need to validate that the packet (and specified UDP length) is large
>>>>>>>>> enough to contain the VXLAN header and the encapsulated packet.
>>>>>>>>>> Bill Fischofer(Bill-Fischofer-Linaro) wrote:
>>>>>>>>>> Inclusion of new APIs means this PR must be against api-next, not
>>>>>>>>>> master.
>>>>>>>>>>> Bill Fischofer(Bill-Fischofer-Linaro) wrote:
>>>>>>>>>>> [RFC 7348](https://tools.ietf.org/html/rfc7348) uses VXLAN, not
>>>>>>>>>>> VxLAN. I think we should use that throughout ODP as well as it
>>>>>>>>>>> avoids unnecessary CamelCase.
>>>>>>>>>>>> Bill Fischofer(Bill-Fischofer-Linaro) wrote:
>>>>>>>>>>>> Same comments here as for helper version of file.
>>>>>>>>>>>>> Bill Fischofer(Bill-Fischofer-Linaro) wrote:
>>>>>>>>>>>>> Typo: reserve, not reseve? Perhaps better for these to be
>>>>>>>>>>>>> anonymous bitfields if they are truly unreferenced?
https://github.com/Linaro/odp/pull/345#discussion_r157219154
updated_at 2017-12-15 14:56:03
