[PATCH][next] mlxsw: spectrum_router: remove redundant initialization of variable force

From: Colin Ian King <redacted>

The variable force is being initialized with a value that is
never read and it is being updated later with a new value. The
initialization is redundant and can be removed.

Addresses-Coverity: ("Unused value")
Signed-off-by: Colin Ian King <redacted>
---
 drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c b/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c
index 6ccaa194733b..ff240e3c29f7 100644
--- a/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c
+++ b/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c

@@ -5059,7 +5059,7 @@ mlxsw_sp_nexthop_obj_bucket_adj_update(struct mlxsw_sp *mlxsw_sp,
 {
 	u16 bucket_index = info->nh_res_bucket->bucket_index;
 	struct netlink_ext_ack *extack = info->extack;
-	bool force = info->nh_res_bucket->force;
+	bool force;
 	char ratr_pl_new[MLXSW_REG_RATR_LEN];
 	char ratr_pl[MLXSW_REG_RATR_LEN];
 	u32 adj_index;

-- 
2.30.2

On Sat, Mar 27, 2021 at 10:33:34PM +0000, Colin King wrote:

Reverse Christmas tree.

regards,
dan carpenter

On Mon, 2021-03-29 at 09:04 +0300, Dan Carpenter wrote:

On Sat, Mar 27, 2021 at 10:33:34PM +0000, Colin King wrote:

quoted

From: Colin Ian King <redacted>

The variable force is being initialized with a value that is
never read and it is being updated later with a new value. The
initialization is redundant and can be removed.

Addresses-Coverity: ("Unused value")
Signed-off-by: Colin Ian King <redacted>
---
 drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c b/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c
index 6ccaa194733b..ff240e3c29f7 100644
--- a/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c
+++ b/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c

@@ -5059,7 +5059,7 @@ mlxsw_sp_nexthop_obj_bucket_adj_update(struct mlxsw_sp *mlxsw_sp,

 {
 	u16 bucket_index = info->nh_res_bucket->bucket_index;
 	struct netlink_ext_ack *extack = info->extack;
-	bool force = info->nh_res_bucket->force;
+	bool force;
 	char ratr_pl_new[MLXSW_REG_RATR_LEN];
 	char ratr_pl[MLXSW_REG_RATR_LEN];
 	u32 adj_index;

Reverse Christmas tree.

Is still terrible style.

On Sat, Mar 27, 2021 at 10:33:34PM +0000, Colin King wrote:

Actually, there is a bug to be fixed here:

diff --git a/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c b/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c
index 6ccaa194733b..41259c0004d1 100644
--- a/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c
+++ b/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c
@@ -5068,8 +5068,9 @@ mlxsw_sp_nexthop_obj_bucket_adj_update(struct mlxsw_sp *mlxsw_sp,
        /* No point in trying an atomic replacement if the idle timer interval
         * is smaller than the interval in which we query and clear activity.
         */
-       force = info->nh_res_bucket->idle_timer_ms <
-               MLXSW_SP_NH_GRP_ACTIVITY_UPDATE_INTERVAL;
+       if (!force && info->nh_res_bucket->idle_timer_ms <
+           MLXSW_SP_NH_GRP_ACTIVITY_UPDATE_INTERVAL)
+               force = true;
 
        adj_index = nh->nhgi->adj_index + bucket_index;
        err = mlxsw_sp_nexthop_update(mlxsw_sp, adj_index, nh, force, ratr_pl);

We should only fallback to a non-atomic replacement when the current
replacement is atomic and the idle timer is too short.

We currently ignore the value of 'force'. This means that a non-atomic
replacement ('force' is true) can be made atomic if idle timer is larger
than 1 second.

Colin, do you mind if I submit it formally as a fix later this week? I
want to run it through our usual process. Will mention you in
Reported-by, obviously.

 	char ratr_pl_new[MLXSW_REG_RATR_LEN];
 	char ratr_pl[MLXSW_REG_RATR_LEN];
 	u32 adj_index;
-- 
2.30.2

On 29/03/2021 08:13, Ido Schimmel wrote:

On Sat, Mar 27, 2021 at 10:33:34PM +0000, Colin King wrote:

quoted

From: Colin Ian King <redacted>

The variable force is being initialized with a value that is
never read and it is being updated later with a new value. The
initialization is redundant and can be removed.

Addresses-Coverity: ("Unused value")
Signed-off-by: Colin Ian King <redacted>
---
 drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c b/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c
index 6ccaa194733b..ff240e3c29f7 100644
--- a/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c
+++ b/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c

@@ -5059,7 +5059,7 @@ mlxsw_sp_nexthop_obj_bucket_adj_update(struct mlxsw_sp *mlxsw_sp,
 {
 	u16 bucket_index = info->nh_res_bucket->bucket_index;
 	struct netlink_ext_ack *extack = info->extack;
-	bool force = info->nh_res_bucket->force;
+	bool force;

Actually, there is a bug to be fixed here:

diff --git a/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c b/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c
index 6ccaa194733b..41259c0004d1 100644
--- a/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c
+++ b/drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c
@@ -5068,8 +5068,9 @@ mlxsw_sp_nexthop_obj_bucket_adj_update(struct mlxsw_sp *mlxsw_sp,
        /* No point in trying an atomic replacement if the idle timer interval
         * is smaller than the interval in which we query and clear activity.
         */
-       force = info->nh_res_bucket->idle_timer_ms <
-               MLXSW_SP_NH_GRP_ACTIVITY_UPDATE_INTERVAL;
+       if (!force && info->nh_res_bucket->idle_timer_ms <
+           MLXSW_SP_NH_GRP_ACTIVITY_UPDATE_INTERVAL)
+               force = true;
 
        adj_index = nh->nhgi->adj_index + bucket_index;
        err = mlxsw_sp_nexthop_update(mlxsw_sp, adj_index, nh, force, ratr_pl);

We should only fallback to a non-atomic replacement when the current
replacement is atomic and the idle timer is too short.

We currently ignore the value of 'force'. This means that a non-atomic
replacement ('force' is true) can be made atomic if idle timer is larger
than 1 second.

Colin, do you mind if I submit it formally as a fix later this week? I
want to run it through our usual process. Will mention you in
Reported-by, obviously.

Sure. Good idea.

quoted

 	char ratr_pl_new[MLXSW_REG_RATR_LEN];
 	char ratr_pl[MLXSW_REG_RATR_LEN];
 	u32 adj_index;
-- 
2.30.2