On Fri, Feb 3, 2023 at 9:49 AM Amit Prakash Shukla
<amitpraka...@marvell.com> wrote:
>
> Initializing rte_graph_param variable with 0 to avoid any
> garbage value in structure elements which are not populated
> as part of this function.
Something looks strange to me.
I fail to see which part of this object is not initialised by the test code.
app/test/test_graph_perf.c: struct rte_graph_param gconf;
app/test/test_graph_perf.c: gconf.socket_id = SOCKET_ID_ANY;
app/test/test_graph_perf.c: gconf.nb_node_patterns = graph_data->nb_nodes;
app/test/test_graph_perf.c: gconf.node_patterns = (const char
**)(uintptr_t)node_patterns;
app/test/test_graph_perf.c: graph_id = rte_graph_create(gname, &gconf);
With this struct defined as:
struct rte_graph_param {
int socket_id; /**< Socket id where memory is allocated. */
uint16_t nb_node_patterns; /**< Number of node patterns. */
const char **node_patterns;
/**< Array of node patterns based on shell pattern. */
};
Is this perhaps related to the following series?
https://patchwork.dpdk.org/project/dpdk/list/?series=26783&state=%2A&archive=both
If this is the case, then this fix should be squashed as part of the
rte_graph_param struct extension.
Thanks.
--
David Marchand
