[PATCH v6 4/8] ethdev: use GRE protocol struct for flow matching

[Thomas Monjalon]

03/02/2023 16:02, Ferruh Yigit:
> On 2/2/2023 5:16 PM, Thomas Monjalon wrote:
> > 02/02/2023 13:44, Ferruh Yigit:
> >> --- a/lib/net/rte_gre.h
> >> +++ b/lib/net/rte_gre.h
> >> @@ -28,6 +28,8 @@ extern "C" {
> >>   */
> >>  __extension__
> >>  struct rte_gre_hdr {
> >> +       union {
> >> +               struct {
> >>  #if RTE_BYTE_ORDER == RTE_LITTLE_ENDIAN
> >>         uint16_t res2:4; /**< Reserved */
> >>         uint16_t s:1;    /**< Sequence Number Present bit */
> >> @@ -45,6 +47,9 @@ struct rte_gre_hdr {
> >>         uint16_t res3:5; /**< Reserved */
> >>         uint16_t ver:3;  /**< Version Number */
> >>  #endif
> >> +               };
> >> +               rte_be16_t c_rsvd0_ver;
> >> +       };
> >>         uint16_t proto;  /**< Protocol Type */
> > 
> > Why adding an unioned field c_rsvd0_ver?
> > 
> > 
> 
> Because existing usage in the drivers require to access these fields as
> a single 16 bytes variable.
> 
> like mlx was using it as:
> `X(SET_BE16,	gre_c_ver, v->c_rsvd0_ver, rte_flow_item_gre) \`
> 
> When all usage switched to flow item specific fields to generic headers,
> there needs a way to represent this in the generic header.
> 
> By adding 'c_rsvd0_ver' to generic header it becomes:
> `X(SET_BE16,	gre_c_ver, v->hdr.c_rsvd0_ver, rte_flow_item_gre) \`
> 
> 
> Or another sample, previous version of code was updated as following:
> `
>  -	size = sizeof(((struct rte_flow_item_gre *)NULL)->c_rsvd0_ver);
>  +	size = sizeof(((struct rte_flow_item_gre *)NULL)->hdr.proto);
> `
> 
> Because generic field to represent 'c_rsvd0_ver' is missing, 'hdr.proto'
> was used, this was wrong.
> Although the sizes of fields are same and functionally works, they are
> different fields, this is worse than sizeof(uint16_t);
> 
> 
> Another usage in testpmd:
> `
>   	[ITEM_GRE_C_RSVD0_VER] = {
>   		.name = "c_rsvd0_ver",
>  @@ -4082,7 +4082,7 @@ static const struct token token_list[] = {
>   		.next = NEXT(item_gre, NEXT_ENTRY(COMMON_UNSIGNED),
>    			     item_param),
>   		.args = ARGS(ARGS_ENTRY_HTON(struct rte_flow_item_gre,
>  -					     c_rsvd0_ver)),
>  +					     hdr.c_rsvd0_ver)),
> `
> 
> 
> But looking it again perhaps it can be named differently, because it is
> not a reserved field in the generic header, though I am not sure what
> can be a good variable name.

The best would be to try not using this field at all.
I suggest to remove this patch from the series for now.
I could try to work on it for the next release.