[PATCH 0/2] net/hns3: bugfix for hns3

[PATCH 0/2] net/hns3: bugfix for hns3

[Jie Hai]

This patchset fixes some bugs.

Dengdui Huang (2):
  net/hns3: fix error code for repeatedly create counter
  net/hns3: fix cannot fully use hardware flow director table

 drivers/net/hns3/hns3_fdir.c | 1 +
 drivers/net/hns3/hns3_flow.c | 2 +-
 2 files changed, 2 insertions(+), 1 deletion(-)

-- 
2.22.0

[PATCH 1/2] net/hns3: fix error code for repeatedly create counter

[Jie Hai]

From: Dengdui Huang <huangdengdui@huawei.com>

Return EINVAL instead of ENOSPC when the same counter ID is
used for multiple times to create a counter.

Fixes: fcba820d9b9e ("net/hns3: support flow director")
Cc: stable@dpdk.org

Signed-off-by: Dengdui Huang <huangdengdui@huawei.com>
Signed-off-by: Jie Hai <haijie1@huawei.com>
---
 drivers/net/hns3/hns3_flow.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/net/hns3/hns3_flow.c b/drivers/net/hns3/hns3_flow.c
index 192ffc015e14..266934b45bce 100644
--- a/drivers/net/hns3/hns3_flow.c
+++ b/drivers/net/hns3/hns3_flow.c
@@ -286,7 +286,7 @@ hns3_counter_new(struct rte_eth_dev *dev, uint32_t indirect, uint32_t id,
 	cnt = hns3_counter_lookup(dev, id);
 	if (cnt) {
 		if (!cnt->indirect || cnt->indirect != indirect)
-			return rte_flow_error_set(error, ENOTSUP,
+			return rte_flow_error_set(error, EINVAL,
 				RTE_FLOW_ERROR_TYPE_ACTION_CONF,
 				cnt,
 				"Counter id is used, indirect flag not match");
-- 
2.22.0

[PATCH 2/2] net/hns3: fix cannot fully use hardware flow director table

[Jie Hai]

From: Dengdui Huang <huangdengdui@huawei.com>

The hns3 driver checks whether the flow rule is repeatedly inserted based
on rte_hash. Currently, the rte_hash extendable bucket table feature is not
enabled. When there are many hash conflicts, the hash table space cannot be
fully used. So the flow rule maybe cannot be inserted even if the hardware
flow director table there are still free. This patch fix it by enabling the
rte_hash extensible bucket table feature.

Fixes: fcba820d9b9e ("net/hns3: support flow director")
Cc: stable@dpdk.org

Signed-off-by: Dengdui Huang <huangdengdui@huawei.com>
Signed-off-by: Jie Hai <haijie1@huawei.com>
---
 drivers/net/hns3/hns3_fdir.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/drivers/net/hns3/hns3_fdir.c b/drivers/net/hns3/hns3_fdir.c
index d18d08353565..aacad40e6174 100644
--- a/drivers/net/hns3/hns3_fdir.c
+++ b/drivers/net/hns3/hns3_fdir.c
@@ -900,6 +900,7 @@ int hns3_fdir_filter_init(struct hns3_adapter *hns)
 		.key_len = sizeof(struct hns3_fdir_key_conf),
 		.hash_func = rte_hash_crc,
 		.hash_func_init_val = 0,
+		.extra_flag = RTE_HASH_EXTRA_FLAGS_EXT_TABLE,
 	};
 	int ret;
 
-- 
2.22.0

Re: [PATCH 2/2] net/hns3: fix cannot fully use hardware flow director table

[Stephen Hemminger]

On Thu, 7 Nov 2024 19:56:45 +0800
Jie Hai <haijie1@huawei.com> wrote:

> From: Dengdui Huang <huangdengdui@huawei.com>
> 
> The hns3 driver checks whether the flow rule is repeatedly inserted based
> on rte_hash. Currently, the rte_hash extendable bucket table feature is not
> enabled. When there are many hash conflicts, the hash table space cannot be
> fully used. So the flow rule maybe cannot be inserted even if the hardware
> flow director table there are still free. This patch fix it by enabling the
> rte_hash extensible bucket table feature.
> 
> Fixes: fcba820d9b9e ("net/hns3: support flow director")
> Cc: stable@dpdk.org
> 
> Signed-off-by: Dengdui Huang <huangdengdui@huawei.com>
> Signed-off-by: Jie Hai <haijie1@huawei.com>
> ---


Acked-by: Stephen Hemminger <stephen@networkplumber.org>

Re: [PATCH 1/2] net/hns3: fix error code for repeatedly create counter

[Stephen Hemminger]

On Thu, 7 Nov 2024 19:56:44 +0800
Jie Hai <haijie1@huawei.com> wrote:

> From: Dengdui Huang <huangdengdui@huawei.com>
> 
> Return EINVAL instead of ENOSPC when the same counter ID is
> used for multiple times to create a counter.
> 
> Fixes: fcba820d9b9e ("net/hns3: support flow director")
> Cc: stable@dpdk.org
> 
> Signed-off-by: Dengdui Huang <huangdengdui@huawei.com>
> Signed-off-by: Jie Hai <haijie1@huawei.com>

Not sure that actual error number matters that much, but
looks good to me.
Another alternative might be EEXIST which is the error code
used when creat() is called on an existing file.

Acked-by: Stephen Hemminger <stephen@networkplumber.org>