Re: [dpdk-dev] [PATCH] lib: fix socket id type

[Stephen Hemminger <stephen@networkplumber.org>]

On Tue, 27 Apr 2021 17:36:32 +0800
"Min Hu (Connor)" <humin29@huawei.com> wrote:

> From: Chengchang Tang <tangchengchang@huawei.com>
> 
> The variable type of the current socket ID is uint8_t. When traversing all
> nodes, the socket ID is compared with RTE_MAX_NUMA_NODES. Since
> RTE_MAX_NUMA_NODES has not been verified, it may be larger than UCHAR_MAX
> theoretically. This would lead to an infinite loop.
> 
> This patch declares the socket ID type as 'int' or 'unsigned int' by
> referring to the common practice in EAL.
> 
> Fixes: 56b6ef874f80 ("efd: new Elastic Flow Distributor library")
> Fixes: f00708c2aa53 ("node: add IPv4 rewrite and lookup control")
> Cc: stable@dpdk.org
> 
> Signed-off-by: Chengchang Tang <tangchengchang@huawei.com>
> Signed-off-by: Min Hu (Connor) <humin29@huawei.com>

An alternative way of handling this would be to add a compile time
check that user did not exceed 255 for numa nodes.

I.e:
diff --git a/lib/efd/rte_efd.c b/lib/efd/rte_efd.c
index dad962ce29bf..8a05909aa4ae 100644
--- a/lib/efd/rte_efd.c
+++ b/lib/efd/rte_efd.c
@@ -505,6 +505,8 @@ rte_efd_create(const char *name, uint32_t max_num_rules, uint32_t key_len,
 	struct rte_ring *r = NULL;
 	unsigned int i;
 
+	RTE_BUILD_BUG_ON(RTE_MAX_NUMA_NODES >= UINT8_MAX);
+
 	efd_list = RTE_TAILQ_CAST(rte_efd_tailq.head, rte_efd_list);
 
 	if (online_cpu_socket_bitmask == 0) {
diff --git a/lib/node/ip4_lookup.c b/lib/node/ip4_lookup.c
index 8bce03d7db9d..0adcd0436b7b 100644
--- a/lib/node/ip4_lookup.c
+++ b/lib/node/ip4_lookup.c
@@ -128,6 +128,8 @@ rte_node_ip4_route_add(uint32_t ip, uint8_t depth, uint16_t next_hop,
 	uint32_t val;
 	int ret;
 
+	RTE_BUILD_BUG_ON(RTE_MAX_NUMA_NODES >= UINT8_MAX);
+
 	in.s_addr = htonl(ip);
 	inet_ntop(AF_INET, &in, abuf, sizeof(abuf));
 	/* Embedded next node id into 24 bit next hop */