Hi Bernard,
<snip>
> +struct rte_flow_classify *
> +rte_flow_classify_create(void *table_handle,
> + uint32_t entry_size,
> + const struct rte_flow_attr *attr,
> + const struct rte_flow_item pattern[],
> + const struct rte_flow_action actions[],
> + struct rte_flow_error *error)
> +{
> + struct rte_flow_classify *flow_classify;
> + int ret;
> +
> + if (!error)
> + return NULL;
> +
> + if (!table_handle) {
> + rte_flow_error_set(error, EINVAL,
> RTE_FLOW_ERROR_TYPE_HANDLE,
> + NULL, "NULL table_handle.");
> + return NULL;
> + }
> +
> + if (!pattern) {
> + rte_flow_error_set(error, EINVAL,
> RTE_FLOW_ERROR_TYPE_ITEM_NUM,
> + NULL, "NULL pattern.");
> + return NULL;
> + }
> +
> + if (!actions) {
> + rte_flow_error_set(error, EINVAL,
> + RTE_FLOW_ERROR_TYPE_ACTION_NUM,
> + NULL, "NULL action.");
> + return NULL;
> + }
> +
> + if (!attr) {
> + rte_flow_error_set(error, EINVAL,
> + RTE_FLOW_ERROR_TYPE_ATTR,
> + NULL, "NULL attribute.");
> + return NULL;
> + }
> +
> + /* parse attr, pattern and actions */
> + ret = rte_flow_classify_validate(table_handle, attr, pattern,
> + actions, error);
> + if (ret < 0)
> + return NULL;
> +
> + flow_classify = allocate_5tuple();
> + if (!flow_classify)
> + return NULL;
> +
> + flow_classify->entry = malloc(entry_size);
> + if (!flow_classify->entry) {
> + free(flow_classify);
> + flow_classify = NULL;
> + return NULL;
> + }
> + memset(flow_classify->entry, 0, entry_size);
> + memmove(flow_classify->entry, &flow_classify->id,
> sizeof(uint32_t));
> +
> + ret = rte_table_acl_ops.f_add(table_handle, &flow_classify-
> >key_add,
> + flow_classify->entry, &flow_classify->key_found,
> + &flow_classify->entry_ptr);
> + if (ret) {
> + free(flow_classify->entry);
> + free(flow_classify);
> + flow_classify = NULL;
> + return NULL;
> + }
> +
> + return flow_classify;
> +}
The API in its current form creates the classifier object which will always use
librte_acl based classification mechanism. This behavior
imposes restriction on the application to always pass only ACL table related
parameters for flow classification. In my opinion, API implementation
should be agnostic to specific classification method and should be generic
enough to allow application to select any of the available flow classification
method (for e.g. acl, hash, LPM, etc.). Otherwise, this library will become
another abstraction of librte_acl for flow classification.
Also, library allows table entries to be added while creating the classifier
object, not later. Is there any specific reason?
