[PATCH net 0/2] DIM fixes for 5.3

[PATCH net 0/2] DIM fixes for 5.3

[Leon Romanovsky]

From: Leon Romanovsky <leonro@mellanox.com>

Hi,

Those two fixes for recently merged DIM patches, both exposed through
RDMa DIM usage.

Thanks

Leon Romanovsky (1):
  lib/dim: Fix -Wunused-const-variable warnings

Yamin Friedman (1):
  linux/dim: Fix overflow in dim calculation

 drivers/net/ethernet/broadcom/bcmsysport.c    |  2 +-
 drivers/net/ethernet/broadcom/bnxt/bnxt.c     |  2 +-
 .../net/ethernet/broadcom/genet/bcmgenet.c    |  2 +-
 .../net/ethernet/mellanox/mlx5/core/en_txrx.c |  4 +-
 include/linux/dim.h                           | 56 -------------------
 lib/dim/dim.c                                 |  4 +-
 lib/dim/net_dim.c                             | 56 +++++++++++++++++++
 7 files changed, 63 insertions(+), 63 deletions(-)

--
2.20.1

[PATCH net 1/2] linux/dim: Fix overflow in dim calculation

[Leon Romanovsky]

From: Yamin Friedman <yaminf@mellanox.com>

While using net_dim, a dim_sample was used without ever initializing the
comps value. Added use of DIV_ROUND_DOWN_ULL() to prevent potential
overflow, it should not be a problem to save the final result in an int
because after the division by epms the value should not be larger than a
few thousand.

[ 1040.127124] UBSAN: Undefined behaviour in lib/dim/dim.c:78:23
[ 1040.130118] signed integer overflow:
[ 1040.131643] 134718714 * 100 cannot be represented in type 'int'

Fixes: 398c2b05bbee ("linux/dim: Add completions count to dim_sample")
Signed-off-by: Yamin Friedman <yaminf@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
---
 drivers/net/ethernet/broadcom/bcmsysport.c        | 2 +-
 drivers/net/ethernet/broadcom/bnxt/bnxt.c         | 2 +-
 drivers/net/ethernet/broadcom/genet/bcmgenet.c    | 2 +-
 drivers/net/ethernet/mellanox/mlx5/core/en_txrx.c | 4 ++--
 lib/dim/dim.c                                     | 4 ++--
 5 files changed, 7 insertions(+), 7 deletions(-)

diff --git a/drivers/net/ethernet/broadcom/bcmsysport.c b/drivers/net/ethernet/broadcom/bcmsysport.c
index b9c5cea8db16..9483553ce444 100644
--- a/drivers/net/ethernet/broadcom/bcmsysport.c
+++ b/drivers/net/ethernet/broadcom/bcmsysport.c
@@ -992,7 +992,7 @@ static int bcm_sysport_poll(struct napi_struct *napi, int budget)
 {
 	struct bcm_sysport_priv *priv =
 		container_of(napi, struct bcm_sysport_priv, napi);
-	struct dim_sample dim_sample;
+	struct dim_sample dim_sample = {};
 	unsigned int work_done = 0;

 	work_done = bcm_sysport_desc_rx(priv, budget);
diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.c b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
index 7134d2c3eb1c..7070349915bc 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt.c
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
@@ -2136,7 +2136,7 @@ static int bnxt_poll(struct napi_struct *napi, int budget)
 		}
 	}
 	if (bp->flags & BNXT_FLAG_DIM) {
-		struct dim_sample dim_sample;
+		struct dim_sample dim_sample = {};

 		dim_update_sample(cpr->event_ctr,
 				  cpr->rx_packets,
diff --git a/drivers/net/ethernet/broadcom/genet/bcmgenet.c b/drivers/net/ethernet/broadcom/genet/bcmgenet.c
index a2b57807453b..d3a0b614dbfa 100644
--- a/drivers/net/ethernet/broadcom/genet/bcmgenet.c
+++ b/drivers/net/ethernet/broadcom/genet/bcmgenet.c
@@ -1895,7 +1895,7 @@ static int bcmgenet_rx_poll(struct napi_struct *napi, int budget)
 {
 	struct bcmgenet_rx_ring *ring = container_of(napi,
 			struct bcmgenet_rx_ring, napi);
-	struct dim_sample dim_sample;
+	struct dim_sample dim_sample = {};
 	unsigned int work_done;

 	work_done = bcmgenet_desc_rx(ring, budget);
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_txrx.c b/drivers/net/ethernet/mellanox/mlx5/core/en_txrx.c
index c50b6f0769c8..49b06b256c92 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_txrx.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_txrx.c
@@ -49,7 +49,7 @@ static inline bool mlx5e_channel_no_affinity_change(struct mlx5e_channel *c)
 static void mlx5e_handle_tx_dim(struct mlx5e_txqsq *sq)
 {
 	struct mlx5e_sq_stats *stats = sq->stats;
-	struct dim_sample dim_sample;
+	struct dim_sample dim_sample = {};

 	if (unlikely(!test_bit(MLX5E_SQ_STATE_AM, &sq->state)))
 		return;
@@ -61,7 +61,7 @@ static void mlx5e_handle_tx_dim(struct mlx5e_txqsq *sq)
 static void mlx5e_handle_rx_dim(struct mlx5e_rq *rq)
 {
 	struct mlx5e_rq_stats *stats = rq->stats;
-	struct dim_sample dim_sample;
+	struct dim_sample dim_sample = {};

 	if (unlikely(!test_bit(MLX5E_RQ_STATE_AM, &rq->state)))
 		return;
diff --git a/lib/dim/dim.c b/lib/dim/dim.c
index 439d641ec796..38045d6d0538 100644
--- a/lib/dim/dim.c
+++ b/lib/dim/dim.c
@@ -74,8 +74,8 @@ void dim_calc_stats(struct dim_sample *start, struct dim_sample *end,
 					delta_us);
 	curr_stats->cpms = DIV_ROUND_UP(ncomps * USEC_PER_MSEC, delta_us);
 	if (curr_stats->epms != 0)
-		curr_stats->cpe_ratio =
-				(curr_stats->cpms * 100) / curr_stats->epms;
+		curr_stats->cpe_ratio = DIV_ROUND_DOWN_ULL(
+			curr_stats->cpms * 100, curr_stats->epms);
 	else
 		curr_stats->cpe_ratio = 0;

--
2.20.1

[PATCH net 2/2] lib/dim: Fix -Wunused-const-variable warnings

[Leon Romanovsky]

From: Leon Romanovsky <leonro@mellanox.com>

DIM causes to the following warnings during kernel compilation
which indicates that tx_profile and rx_profile are supposed to
be declared in *.c and not in *.h files.

In file included from ./include/rdma/ib_verbs.h:64,
                 from ./include/linux/mlx5/device.h:37,
                 from ./include/linux/mlx5/driver.h:51,
                 from ./include/linux/mlx5/vport.h:36,
                 from drivers/infiniband/hw/mlx5/ib_virt.c:34:
./include/linux/dim.h:326:1: warning: _tx_profile_ defined but not used [-Wunused-const-variable=]
  326 | tx_profile[DIM_CQ_PERIOD_NUM_MODES][NET_DIM_PARAMS_NUM_PROFILES] = {
      | ^~~~~~~~~~
./include/linux/dim.h:320:1: warning: _rx_profile_ defined but not used [-Wunused-const-variable=]
  320 | rx_profile[DIM_CQ_PERIOD_NUM_MODES][NET_DIM_PARAMS_NUM_PROFILES] = {
      | ^~~~~~~~~~

Fixes: 4f75da3666c0 ("linux/dim: Move implementation to .c files")
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
---
 include/linux/dim.h | 56 ---------------------------------------------
 lib/dim/net_dim.c   | 56 +++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 56 insertions(+), 56 deletions(-)

diff --git a/include/linux/dim.h b/include/linux/dim.h
index d3a0fbfff2bb..9fa4b3f88c39 100644
--- a/include/linux/dim.h
+++ b/include/linux/dim.h
@@ -272,62 +272,6 @@ dim_update_sample_with_comps(u16 event_ctr, u64 packets, u64 bytes, u64 comps,

 /* Net DIM */

-/*
- * Net DIM profiles:
- *        There are different set of profiles for each CQ period mode.
- *        There are different set of profiles for RX/TX CQs.
- *        Each profile size must be of NET_DIM_PARAMS_NUM_PROFILES
- */
-#define NET_DIM_PARAMS_NUM_PROFILES 5
-#define NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE 256
-#define NET_DIM_DEFAULT_TX_CQ_MODERATION_PKTS_FROM_EQE 128
-#define NET_DIM_DEF_PROFILE_CQE 1
-#define NET_DIM_DEF_PROFILE_EQE 1
-
-#define NET_DIM_RX_EQE_PROFILES { \
-	{1,   NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \
-	{8,   NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \
-	{64,  NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \
-	{128, NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \
-	{256, NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \
-}
-
-#define NET_DIM_RX_CQE_PROFILES { \
-	{2,  256},             \
-	{8,  128},             \
-	{16, 64},              \
-	{32, 64},              \
-	{64, 64}               \
-}
-
-#define NET_DIM_TX_EQE_PROFILES { \
-	{1,   NET_DIM_DEFAULT_TX_CQ_MODERATION_PKTS_FROM_EQE},  \
-	{8,   NET_DIM_DEFAULT_TX_CQ_MODERATION_PKTS_FROM_EQE},  \
-	{32,  NET_DIM_DEFAULT_TX_CQ_MODERATION_PKTS_FROM_EQE},  \
-	{64,  NET_DIM_DEFAULT_TX_CQ_MODERATION_PKTS_FROM_EQE},  \
-	{128, NET_DIM_DEFAULT_TX_CQ_MODERATION_PKTS_FROM_EQE}   \
-}
-
-#define NET_DIM_TX_CQE_PROFILES { \
-	{5,  128},  \
-	{8,  64},  \
-	{16, 32},  \
-	{32, 32},  \
-	{64, 32}   \
-}
-
-static const struct dim_cq_moder
-rx_profile[DIM_CQ_PERIOD_NUM_MODES][NET_DIM_PARAMS_NUM_PROFILES] = {
-	NET_DIM_RX_EQE_PROFILES,
-	NET_DIM_RX_CQE_PROFILES,
-};
-
-static const struct dim_cq_moder
-tx_profile[DIM_CQ_PERIOD_NUM_MODES][NET_DIM_PARAMS_NUM_PROFILES] = {
-	NET_DIM_TX_EQE_PROFILES,
-	NET_DIM_TX_CQE_PROFILES,
-};
-
 /**
  *	net_dim_get_rx_moderation - provide a CQ moderation object for the given RX profile
  *	@cq_period_mode: CQ period mode
diff --git a/lib/dim/net_dim.c b/lib/dim/net_dim.c
index 5bcc902c5388..a4db51c21266 100644
--- a/lib/dim/net_dim.c
+++ b/lib/dim/net_dim.c
@@ -5,6 +5,62 @@

 #include <linux/dim.h>

+/*
+ * Net DIM profiles:
+ *        There are different set of profiles for each CQ period mode.
+ *        There are different set of profiles for RX/TX CQs.
+ *        Each profile size must be of NET_DIM_PARAMS_NUM_PROFILES
+ */
+#define NET_DIM_PARAMS_NUM_PROFILES 5
+#define NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE 256
+#define NET_DIM_DEFAULT_TX_CQ_MODERATION_PKTS_FROM_EQE 128
+#define NET_DIM_DEF_PROFILE_CQE 1
+#define NET_DIM_DEF_PROFILE_EQE 1
+
+#define NET_DIM_RX_EQE_PROFILES { \
+	{1,   NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \
+	{8,   NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \
+	{64,  NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \
+	{128, NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \
+	{256, NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \
+}
+
+#define NET_DIM_RX_CQE_PROFILES { \
+	{2,  256},             \
+	{8,  128},             \
+	{16, 64},              \
+	{32, 64},              \
+	{64, 64}               \
+}
+
+#define NET_DIM_TX_EQE_PROFILES { \
+	{1,   NET_DIM_DEFAULT_TX_CQ_MODERATION_PKTS_FROM_EQE},  \
+	{8,   NET_DIM_DEFAULT_TX_CQ_MODERATION_PKTS_FROM_EQE},  \
+	{32,  NET_DIM_DEFAULT_TX_CQ_MODERATION_PKTS_FROM_EQE},  \
+	{64,  NET_DIM_DEFAULT_TX_CQ_MODERATION_PKTS_FROM_EQE},  \
+	{128, NET_DIM_DEFAULT_TX_CQ_MODERATION_PKTS_FROM_EQE}   \
+}
+
+#define NET_DIM_TX_CQE_PROFILES { \
+	{5,  128},  \
+	{8,  64},  \
+	{16, 32},  \
+	{32, 32},  \
+	{64, 32}   \
+}
+
+static const struct dim_cq_moder
+rx_profile[DIM_CQ_PERIOD_NUM_MODES][NET_DIM_PARAMS_NUM_PROFILES] = {
+	NET_DIM_RX_EQE_PROFILES,
+	NET_DIM_RX_CQE_PROFILES,
+};
+
+static const struct dim_cq_moder
+tx_profile[DIM_CQ_PERIOD_NUM_MODES][NET_DIM_PARAMS_NUM_PROFILES] = {
+	NET_DIM_TX_EQE_PROFILES,
+	NET_DIM_TX_CQE_PROFILES,
+};
+
 struct dim_cq_moder
 net_dim_get_rx_moderation(u8 cq_period_mode, int ix)
 {
--
2.20.1

Re: [PATCH net 2/2] lib/dim: Fix -Wunused-const-variable warnings

[Bart Van Assche]

On 7/23/19 12:22 AM, Leon Romanovsky wrote:
> From: Leon Romanovsky <leonro@mellanox.com>
> 
> DIM causes to the following warnings during kernel compilation
> which indicates that tx_profile and rx_profile are supposed to
> be declared in *.c and not in *.h files.

Thanks Leon for this fix.

Reviewed-by: Bart Van Assche <bvanassche@acm.org>

Re: [PATCH net 2/2] lib/dim: Fix -Wunused-const-variable warnings

[Saeed Mahameed]

On Tue, 2019-07-23 at 10:22 +0300, Leon Romanovsky wrote:
> From: Leon Romanovsky <leonro@mellanox.com>
> 
> DIM causes to the following warnings during kernel compilation
> which indicates that tx_profile and rx_profile are supposed to
> be declared in *.c and not in *.h files.
> 
> In file included from ./include/rdma/ib_verbs.h:64,
>                  from ./include/linux/mlx5/device.h:37,
>                  from ./include/linux/mlx5/driver.h:51,
>                  from ./include/linux/mlx5/vport.h:36,
>                  from drivers/infiniband/hw/mlx5/ib_virt.c:34:
> ./include/linux/dim.h:326:1: warning: _tx_profile_ defined but not
> used [-Wunused-const-variable=]
>   326 |
> tx_profile[DIM_CQ_PERIOD_NUM_MODES][NET_DIM_PARAMS_NUM_PROFILES] = {
>       | ^~~~~~~~~~
> ./include/linux/dim.h:320:1: warning: _rx_profile_ defined but not
> used [-Wunused-const-variable=]
>   320 |
> rx_profile[DIM_CQ_PERIOD_NUM_MODES][NET_DIM_PARAMS_NUM_PROFILES] = {
>       | ^~~~~~~~~~
> 
> Fixes: 4f75da3666c0 ("linux/dim: Move implementation to .c files")
> Signed-off-by: Leon Romanovsky <leonro@mellanox.com>

A similar patch was already submitted to linux-kernel ML, 
"[PATCH] linux/dim: fix -Wunused-const-variable warnings"

I basically asked Qian to do the same as you did in this patch.
Anyway i guess it is ok to drop that patch and keep this one.

Acked-by: Saeed Mahameed <saeedm@mellanox.com>

> ---
>  include/linux/dim.h | 56 -----------------------------------------
> ----
>  lib/dim/net_dim.c   | 56
> +++++++++++++++++++++++++++++++++++++++++++++
>  2 files changed, 56 insertions(+), 56 deletions(-)
> 
> diff --git a/include/linux/dim.h b/include/linux/dim.h
> index d3a0fbfff2bb..9fa4b3f88c39 100644
> --- a/include/linux/dim.h
> +++ b/include/linux/dim.h
> @@ -272,62 +272,6 @@ dim_update_sample_with_comps(u16 event_ctr, u64
> packets, u64 bytes, u64 comps,
> 
>  /* Net DIM */
> 
> -/*
> - * Net DIM profiles:
> - *        There are different set of profiles for each CQ period
> mode.
> - *        There are different set of profiles for RX/TX CQs.
> - *        Each profile size must be of NET_DIM_PARAMS_NUM_PROFILES
> - */
> -#define NET_DIM_PARAMS_NUM_PROFILES 5
> -#define NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE 256
> -#define NET_DIM_DEFAULT_TX_CQ_MODERATION_PKTS_FROM_EQE 128
> -#define NET_DIM_DEF_PROFILE_CQE 1
> -#define NET_DIM_DEF_PROFILE_EQE 1
> -
> -#define NET_DIM_RX_EQE_PROFILES { \
> -	{1,   NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \
> -	{8,   NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \
> -	{64,  NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \
> -	{128, NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \
> -	{256, NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \
> -}
> -
> -#define NET_DIM_RX_CQE_PROFILES { \
> -	{2,  256},             \
> -	{8,  128},             \
> -	{16, 64},              \
> -	{32, 64},              \
> -	{64, 64}               \
> -}
> -
> -#define NET_DIM_TX_EQE_PROFILES { \
> -	{1,   NET_DIM_DEFAULT_TX_CQ_MODERATION_PKTS_FROM_EQE},  \
> -	{8,   NET_DIM_DEFAULT_TX_CQ_MODERATION_PKTS_FROM_EQE},  \
> -	{32,  NET_DIM_DEFAULT_TX_CQ_MODERATION_PKTS_FROM_EQE},  \
> -	{64,  NET_DIM_DEFAULT_TX_CQ_MODERATION_PKTS_FROM_EQE},  \
> -	{128, NET_DIM_DEFAULT_TX_CQ_MODERATION_PKTS_FROM_EQE}   \
> -}
> -
> -#define NET_DIM_TX_CQE_PROFILES { \
> -	{5,  128},  \
> -	{8,  64},  \
> -	{16, 32},  \
> -	{32, 32},  \
> -	{64, 32}   \
> -}
> -
> -static const struct dim_cq_moder
> -rx_profile[DIM_CQ_PERIOD_NUM_MODES][NET_DIM_PARAMS_NUM_PROFILES] = {
> -	NET_DIM_RX_EQE_PROFILES,
> -	NET_DIM_RX_CQE_PROFILES,
> -};
> -
> -static const struct dim_cq_moder
> -tx_profile[DIM_CQ_PERIOD_NUM_MODES][NET_DIM_PARAMS_NUM_PROFILES] = {
> -	NET_DIM_TX_EQE_PROFILES,
> -	NET_DIM_TX_CQE_PROFILES,
> -};
> -
>  /**
>   *	net_dim_get_rx_moderation - provide a CQ moderation object for
> the given RX profile
>   *	@cq_period_mode: CQ period mode
> diff --git a/lib/dim/net_dim.c b/lib/dim/net_dim.c
> index 5bcc902c5388..a4db51c21266 100644
> --- a/lib/dim/net_dim.c
> +++ b/lib/dim/net_dim.c
> @@ -5,6 +5,62 @@
> 
>  #include <linux/dim.h>
> 
> +/*
> + * Net DIM profiles:
> + *        There are different set of profiles for each CQ period
> mode.
> + *        There are different set of profiles for RX/TX CQs.
> + *        Each profile size must be of NET_DIM_PARAMS_NUM_PROFILES
> + */
> +#define NET_DIM_PARAMS_NUM_PROFILES 5
> +#define NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE 256
> +#define NET_DIM_DEFAULT_TX_CQ_MODERATION_PKTS_FROM_EQE 128
> +#define NET_DIM_DEF_PROFILE_CQE 1
> +#define NET_DIM_DEF_PROFILE_EQE 1
> +
> +#define NET_DIM_RX_EQE_PROFILES { \
> +	{1,   NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \
> +	{8,   NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \
> +	{64,  NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \
> +	{128, NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \
> +	{256, NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \
> +}
> +
> +#define NET_DIM_RX_CQE_PROFILES { \
> +	{2,  256},             \
> +	{8,  128},             \
> +	{16, 64},              \
> +	{32, 64},              \
> +	{64, 64}               \
> +}
> +
> +#define NET_DIM_TX_EQE_PROFILES { \
> +	{1,   NET_DIM_DEFAULT_TX_CQ_MODERATION_PKTS_FROM_EQE},  \
> +	{8,   NET_DIM_DEFAULT_TX_CQ_MODERATION_PKTS_FROM_EQE},  \
> +	{32,  NET_DIM_DEFAULT_TX_CQ_MODERATION_PKTS_FROM_EQE},  \
> +	{64,  NET_DIM_DEFAULT_TX_CQ_MODERATION_PKTS_FROM_EQE},  \
> +	{128, NET_DIM_DEFAULT_TX_CQ_MODERATION_PKTS_FROM_EQE}   \
> +}
> +
> +#define NET_DIM_TX_CQE_PROFILES { \
> +	{5,  128},  \
> +	{8,  64},  \
> +	{16, 32},  \
> +	{32, 32},  \
> +	{64, 32}   \
> +}
> +
> +static const struct dim_cq_moder
> +rx_profile[DIM_CQ_PERIOD_NUM_MODES][NET_DIM_PARAMS_NUM_PROFILES] = {
> +	NET_DIM_RX_EQE_PROFILES,
> +	NET_DIM_RX_CQE_PROFILES,
> +};
> +
> +static const struct dim_cq_moder
> +tx_profile[DIM_CQ_PERIOD_NUM_MODES][NET_DIM_PARAMS_NUM_PROFILES] = {
> +	NET_DIM_TX_EQE_PROFILES,
> +	NET_DIM_TX_CQE_PROFILES,
> +};
> +
>  struct dim_cq_moder
>  net_dim_get_rx_moderation(u8 cq_period_mode, int ix)
>  {
> --
> 2.20.1
> 

Re: [PATCH net 1/2] linux/dim: Fix overflow in dim calculation

[Saeed Mahameed]

On Tue, 2019-07-23 at 10:22 +0300, Leon Romanovsky wrote:
> From: Yamin Friedman <yaminf@mellanox.com>
> 
> While using net_dim, a dim_sample was used without ever initializing
> the
> comps value. Added use of DIV_ROUND_DOWN_ULL() to prevent potential
> overflow, it should not be a problem to save the final result in an
> int
> because after the division by epms the value should not be larger
> than a
> few thousand.
> 
> [ 1040.127124] UBSAN: Undefined behaviour in lib/dim/dim.c:78:23
> [ 1040.130118] signed integer overflow:
> [ 1040.131643] 134718714 * 100 cannot be represented in type 'int'
> 
> Fixes: 398c2b05bbee ("linux/dim: Add completions count to
> dim_sample")
> Signed-off-by: Yamin Friedman <yaminf@mellanox.com>
> Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
> ---
>  drivers/net/ethernet/broadcom/bcmsysport.c        | 2 +-
>  drivers/net/ethernet/broadcom/bnxt/bnxt.c         | 2 +-
>  drivers/net/ethernet/broadcom/genet/bcmgenet.c    | 2 +-
>  drivers/net/ethernet/mellanox/mlx5/core/en_txrx.c | 4 ++--
>  lib/dim/dim.c                                     | 4 ++--
>  5 files changed, 7 insertions(+), 7 deletions(-)
> 
> diff --git a/drivers/net/ethernet/broadcom/bcmsysport.c
> b/drivers/net/ethernet/broadcom/bcmsysport.c
> index b9c5cea8db16..9483553ce444 100644
> --- a/drivers/net/ethernet/broadcom/bcmsysport.c
> +++ b/drivers/net/ethernet/broadcom/bcmsysport.c
> @@ -992,7 +992,7 @@ static int bcm_sysport_poll(struct napi_struct
> *napi, int budget)
>  {
>  	struct bcm_sysport_priv *priv =
>  		container_of(napi, struct bcm_sysport_priv, napi);
> -	struct dim_sample dim_sample;
> +	struct dim_sample dim_sample = {};

net_dim implementation doesn't care about sample->comp_ctr, so this is
unnecessary for the sake of fixing the rdma overflow issue, but it
doens't hurt anyone to have this change in this patch, although Tariq
already sent me a fix that i applied to my internal queues.

>  	unsigned int work_done = 0;
> 
>  	work_done = bcm_sysport_desc_rx(priv, budget);
> diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.c
> b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
> index 7134d2c3eb1c..7070349915bc 100644
> --- a/drivers/net/ethernet/broadcom/bnxt/bnxt.c
> +++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
> @@ -2136,7 +2136,7 @@ static int bnxt_poll(struct napi_struct *napi,
> int budget)
>  		}
>  	}
>  	if (bp->flags & BNXT_FLAG_DIM) {
> -		struct dim_sample dim_sample;
> +		struct dim_sample dim_sample = {};
> 
>  		dim_update_sample(cpr->event_ctr,
>  				  cpr->rx_packets,
> diff --git a/drivers/net/ethernet/broadcom/genet/bcmgenet.c
> b/drivers/net/ethernet/broadcom/genet/bcmgenet.c
> index a2b57807453b..d3a0b614dbfa 100644
> --- a/drivers/net/ethernet/broadcom/genet/bcmgenet.c
> +++ b/drivers/net/ethernet/broadcom/genet/bcmgenet.c
> @@ -1895,7 +1895,7 @@ static int bcmgenet_rx_poll(struct napi_struct
> *napi, int budget)
>  {
>  	struct bcmgenet_rx_ring *ring = container_of(napi,
>  			struct bcmgenet_rx_ring, napi);
> -	struct dim_sample dim_sample;
> +	struct dim_sample dim_sample = {};
>  	unsigned int work_done;
> 
>  	work_done = bcmgenet_desc_rx(ring, budget);
> diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_txrx.c
> b/drivers/net/ethernet/mellanox/mlx5/core/en_txrx.c
> index c50b6f0769c8..49b06b256c92 100644
> --- a/drivers/net/ethernet/mellanox/mlx5/core/en_txrx.c
> +++ b/drivers/net/ethernet/mellanox/mlx5/core/en_txrx.c
> @@ -49,7 +49,7 @@ static inline bool
> mlx5e_channel_no_affinity_change(struct mlx5e_channel *c)
>  static void mlx5e_handle_tx_dim(struct mlx5e_txqsq *sq)
>  {
>  	struct mlx5e_sq_stats *stats = sq->stats;
> -	struct dim_sample dim_sample;
> +	struct dim_sample dim_sample = {};
> 
>  	if (unlikely(!test_bit(MLX5E_SQ_STATE_AM, &sq->state)))
>  		return;
> @@ -61,7 +61,7 @@ static void mlx5e_handle_tx_dim(struct mlx5e_txqsq
> *sq)
>  static void mlx5e_handle_rx_dim(struct mlx5e_rq *rq)
>  {
>  	struct mlx5e_rq_stats *stats = rq->stats;
> -	struct dim_sample dim_sample;
> +	struct dim_sample dim_sample = {};
> 
>  	if (unlikely(!test_bit(MLX5E_RQ_STATE_AM, &rq->state)))
>  		return;
> diff --git a/lib/dim/dim.c b/lib/dim/dim.c
> index 439d641ec796..38045d6d0538 100644
> --- a/lib/dim/dim.c
> +++ b/lib/dim/dim.c
> @@ -74,8 +74,8 @@ void dim_calc_stats(struct dim_sample *start,
> struct dim_sample *end,
>  					delta_us);
>  	curr_stats->cpms = DIV_ROUND_UP(ncomps * USEC_PER_MSEC,
> delta_us);
>  	if (curr_stats->epms != 0)

BTW unrelated to this changed but curr_stats->epms  can never be 0 due
to DIV_ROUND_UP. we can save a condition here.

> -		curr_stats->cpe_ratio =
> -				(curr_stats->cpms * 100) / curr_stats-
> >epms;
> +		curr_stats->cpe_ratio = DIV_ROUND_DOWN_ULL(
> +			curr_stats->cpms * 100, curr_stats->epms);
>  	else
>  		curr_stats->cpe_ratio = 0;
> 

LGTM,

Acked-by: Saeed Mahameed <saeedm@mellanox.com>

> --
> 2.20.1
> 

Re: [PATCH net 2/2] lib/dim: Fix -Wunused-const-variable warnings

[Leon Romanovsky]

On Tue, Jul 23, 2019 at 05:05:19PM +0000, Saeed Mahameed wrote:
> On Tue, 2019-07-23 at 10:22 +0300, Leon Romanovsky wrote:
> > From: Leon Romanovsky <leonro@mellanox.com>
> >
> > DIM causes to the following warnings during kernel compilation
> > which indicates that tx_profile and rx_profile are supposed to
> > be declared in *.c and not in *.h files.
> >
> > In file included from ./include/rdma/ib_verbs.h:64,
> >                  from ./include/linux/mlx5/device.h:37,
> >                  from ./include/linux/mlx5/driver.h:51,
> >                  from ./include/linux/mlx5/vport.h:36,
> >                  from drivers/infiniband/hw/mlx5/ib_virt.c:34:
> > ./include/linux/dim.h:326:1: warning: _tx_profile_ defined but not
> > used [-Wunused-const-variable=]
> >   326 |
> > tx_profile[DIM_CQ_PERIOD_NUM_MODES][NET_DIM_PARAMS_NUM_PROFILES] = {
> >       | ^~~~~~~~~~
> > ./include/linux/dim.h:320:1: warning: _rx_profile_ defined but not
> > used [-Wunused-const-variable=]
> >   320 |
> > rx_profile[DIM_CQ_PERIOD_NUM_MODES][NET_DIM_PARAMS_NUM_PROFILES] = {
> >       | ^~~~~~~~~~
> >
> > Fixes: 4f75da3666c0 ("linux/dim: Move implementation to .c files")
> > Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
>
> A similar patch was already submitted to linux-kernel ML,
> "[PATCH] linux/dim: fix -Wunused-const-variable warnings"

Are you talking about this merged patch? If yes, it was incomplete.

ommit bedc0fd0f9b517698193d644f914b33951856fd2
Author: Qian Cai <cai@lca.pw>
Date:   Thu Jul 11 09:55:56 2019 -0400

    RDMA/core: Fix -Wunused-const-variable warnings


>
> I basically asked Qian to do the same as you did in this patch.
> Anyway i guess it is ok to drop that patch and keep this one.
>
> Acked-by: Saeed Mahameed <saeedm@mellanox.com>
>
> > ---
> >  include/linux/dim.h | 56 -----------------------------------------
> > ----
> >  lib/dim/net_dim.c   | 56
> > +++++++++++++++++++++++++++++++++++++++++++++
> >  2 files changed, 56 insertions(+), 56 deletions(-)
> >
> > diff --git a/include/linux/dim.h b/include/linux/dim.h
> > index d3a0fbfff2bb..9fa4b3f88c39 100644
> > --- a/include/linux/dim.h
> > +++ b/include/linux/dim.h
> > @@ -272,62 +272,6 @@ dim_update_sample_with_comps(u16 event_ctr, u64
> > packets, u64 bytes, u64 comps,
> >
> >  /* Net DIM */
> >
> > -/*
> > - * Net DIM profiles:
> > - *        There are different set of profiles for each CQ period
> > mode.
> > - *        There are different set of profiles for RX/TX CQs.
> > - *        Each profile size must be of NET_DIM_PARAMS_NUM_PROFILES
> > - */
> > -#define NET_DIM_PARAMS_NUM_PROFILES 5
> > -#define NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE 256
> > -#define NET_DIM_DEFAULT_TX_CQ_MODERATION_PKTS_FROM_EQE 128
> > -#define NET_DIM_DEF_PROFILE_CQE 1
> > -#define NET_DIM_DEF_PROFILE_EQE 1
> > -
> > -#define NET_DIM_RX_EQE_PROFILES { \
> > -	{1,   NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \
> > -	{8,   NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \
> > -	{64,  NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \
> > -	{128, NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \
> > -	{256, NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \
> > -}
> > -
> > -#define NET_DIM_RX_CQE_PROFILES { \
> > -	{2,  256},             \
> > -	{8,  128},             \
> > -	{16, 64},              \
> > -	{32, 64},              \
> > -	{64, 64}               \
> > -}
> > -
> > -#define NET_DIM_TX_EQE_PROFILES { \
> > -	{1,   NET_DIM_DEFAULT_TX_CQ_MODERATION_PKTS_FROM_EQE},  \
> > -	{8,   NET_DIM_DEFAULT_TX_CQ_MODERATION_PKTS_FROM_EQE},  \
> > -	{32,  NET_DIM_DEFAULT_TX_CQ_MODERATION_PKTS_FROM_EQE},  \
> > -	{64,  NET_DIM_DEFAULT_TX_CQ_MODERATION_PKTS_FROM_EQE},  \
> > -	{128, NET_DIM_DEFAULT_TX_CQ_MODERATION_PKTS_FROM_EQE}   \
> > -}
> > -
> > -#define NET_DIM_TX_CQE_PROFILES { \
> > -	{5,  128},  \
> > -	{8,  64},  \
> > -	{16, 32},  \
> > -	{32, 32},  \
> > -	{64, 32}   \
> > -}
> > -
> > -static const struct dim_cq_moder
> > -rx_profile[DIM_CQ_PERIOD_NUM_MODES][NET_DIM_PARAMS_NUM_PROFILES] = {
> > -	NET_DIM_RX_EQE_PROFILES,
> > -	NET_DIM_RX_CQE_PROFILES,
> > -};
> > -
> > -static const struct dim_cq_moder
> > -tx_profile[DIM_CQ_PERIOD_NUM_MODES][NET_DIM_PARAMS_NUM_PROFILES] = {
> > -	NET_DIM_TX_EQE_PROFILES,
> > -	NET_DIM_TX_CQE_PROFILES,
> > -};
> > -
> >  /**
> >   *	net_dim_get_rx_moderation - provide a CQ moderation object for
> > the given RX profile
> >   *	@cq_period_mode: CQ period mode
> > diff --git a/lib/dim/net_dim.c b/lib/dim/net_dim.c
> > index 5bcc902c5388..a4db51c21266 100644
> > --- a/lib/dim/net_dim.c
> > +++ b/lib/dim/net_dim.c
> > @@ -5,6 +5,62 @@
> >
> >  #include <linux/dim.h>
> >
> > +/*
> > + * Net DIM profiles:
> > + *        There are different set of profiles for each CQ period
> > mode.
> > + *        There are different set of profiles for RX/TX CQs.
> > + *        Each profile size must be of NET_DIM_PARAMS_NUM_PROFILES
> > + */
> > +#define NET_DIM_PARAMS_NUM_PROFILES 5
> > +#define NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE 256
> > +#define NET_DIM_DEFAULT_TX_CQ_MODERATION_PKTS_FROM_EQE 128
> > +#define NET_DIM_DEF_PROFILE_CQE 1
> > +#define NET_DIM_DEF_PROFILE_EQE 1
> > +
> > +#define NET_DIM_RX_EQE_PROFILES { \
> > +	{1,   NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \
> > +	{8,   NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \
> > +	{64,  NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \
> > +	{128, NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \
> > +	{256, NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \
> > +}
> > +
> > +#define NET_DIM_RX_CQE_PROFILES { \
> > +	{2,  256},             \
> > +	{8,  128},             \
> > +	{16, 64},              \
> > +	{32, 64},              \
> > +	{64, 64}               \
> > +}
> > +
> > +#define NET_DIM_TX_EQE_PROFILES { \
> > +	{1,   NET_DIM_DEFAULT_TX_CQ_MODERATION_PKTS_FROM_EQE},  \
> > +	{8,   NET_DIM_DEFAULT_TX_CQ_MODERATION_PKTS_FROM_EQE},  \
> > +	{32,  NET_DIM_DEFAULT_TX_CQ_MODERATION_PKTS_FROM_EQE},  \
> > +	{64,  NET_DIM_DEFAULT_TX_CQ_MODERATION_PKTS_FROM_EQE},  \
> > +	{128, NET_DIM_DEFAULT_TX_CQ_MODERATION_PKTS_FROM_EQE}   \
> > +}
> > +
> > +#define NET_DIM_TX_CQE_PROFILES { \
> > +	{5,  128},  \
> > +	{8,  64},  \
> > +	{16, 32},  \
> > +	{32, 32},  \
> > +	{64, 32}   \
> > +}
> > +
> > +static const struct dim_cq_moder
> > +rx_profile[DIM_CQ_PERIOD_NUM_MODES][NET_DIM_PARAMS_NUM_PROFILES] = {
> > +	NET_DIM_RX_EQE_PROFILES,
> > +	NET_DIM_RX_CQE_PROFILES,
> > +};
> > +
> > +static const struct dim_cq_moder
> > +tx_profile[DIM_CQ_PERIOD_NUM_MODES][NET_DIM_PARAMS_NUM_PROFILES] = {
> > +	NET_DIM_TX_EQE_PROFILES,
> > +	NET_DIM_TX_CQE_PROFILES,
> > +};
> > +
> >  struct dim_cq_moder
> >  net_dim_get_rx_moderation(u8 cq_period_mode, int ix)
> >  {
> > --
> > 2.20.1
> >

Re: [PATCH net 1/2] linux/dim: Fix overflow in dim calculation

[Leon Romanovsky]

On Tue, Jul 23, 2019 at 05:22:43PM +0000, Saeed Mahameed wrote:
> On Tue, 2019-07-23 at 10:22 +0300, Leon Romanovsky wrote:
> > From: Yamin Friedman <yaminf@mellanox.com>
> >
> > While using net_dim, a dim_sample was used without ever initializing
> > the
> > comps value. Added use of DIV_ROUND_DOWN_ULL() to prevent potential
> > overflow, it should not be a problem to save the final result in an
> > int
> > because after the division by epms the value should not be larger
> > than a
> > few thousand.
> >
> > [ 1040.127124] UBSAN: Undefined behaviour in lib/dim/dim.c:78:23
> > [ 1040.130118] signed integer overflow:
> > [ 1040.131643] 134718714 * 100 cannot be represented in type 'int'
> >
> > Fixes: 398c2b05bbee ("linux/dim: Add completions count to
> > dim_sample")
> > Signed-off-by: Yamin Friedman <yaminf@mellanox.com>
> > Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
> > ---
> >  drivers/net/ethernet/broadcom/bcmsysport.c        | 2 +-
> >  drivers/net/ethernet/broadcom/bnxt/bnxt.c         | 2 +-
> >  drivers/net/ethernet/broadcom/genet/bcmgenet.c    | 2 +-
> >  drivers/net/ethernet/mellanox/mlx5/core/en_txrx.c | 4 ++--
> >  lib/dim/dim.c                                     | 4 ++--
> >  5 files changed, 7 insertions(+), 7 deletions(-)
> >
> > diff --git a/drivers/net/ethernet/broadcom/bcmsysport.c
> > b/drivers/net/ethernet/broadcom/bcmsysport.c
> > index b9c5cea8db16..9483553ce444 100644
> > --- a/drivers/net/ethernet/broadcom/bcmsysport.c
> > +++ b/drivers/net/ethernet/broadcom/bcmsysport.c
> > @@ -992,7 +992,7 @@ static int bcm_sysport_poll(struct napi_struct
> > *napi, int budget)
> >  {
> >  	struct bcm_sysport_priv *priv =
> >  		container_of(napi, struct bcm_sysport_priv, napi);
> > -	struct dim_sample dim_sample;
> > +	struct dim_sample dim_sample = {};
>
> net_dim implementation doesn't care about sample->comp_ctr, so this is
> unnecessary for the sake of fixing the rdma overflow issue, but it
> doens't hurt anyone to have this change in this patch.

Yes, this is why we decided to change all drivers and not mlx5 only.

Thanks

Re: [PATCH net 0/2] DIM fixes for 5.3

[David Miller]

From: Leon Romanovsky <leon@kernel.org>
Date: Tue, 23 Jul 2019 10:22:46 +0300

> Those two fixes for recently merged DIM patches, both exposed through
> RDMa DIM usage.

Series applied.