[pull request][net-next 00/12] Mellanox, mlx5 updates 2019-08-01

[pull request][net-next 00/12] Mellanox, mlx5 updates 2019-08-01

[Saeed Mahameed]

Hi Dave,

This series provides misc updates to mlx5 driver.
For more information please see tag log below.

Please pull and let me know if there is any problem.

Please note that the series starts with a merge of mlx5-next branch,
to resolve and avoid dependency with rdma tree.

Thanks,
Saeed.

---
The following changes since commit 68e18626dfe971df3856872ee58f63c389dea2f5:

  Merge branch 'mlx5-next' of git://git.kernel.org/pub/scm/linux/kernel/git/mellanox/linux (2019-08-01 12:33:14 -0700)

are available in the Git repository at:

  git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linux.git tags/mlx5-updates-2019-08-01

for you to fetch changes up to 6830b468259b45e3b73070474b8cec9388aa8c11:

  net/mlx5e: Allow dropping specific tunnel packets (2019-08-01 12:33:33 -0700)

----------------------------------------------------------------
mlx5-updates-2019-08-01

Misc updates for mlx5 netdev driver:

1) Ingress rate support for E-Switch vports from Eli.
2) Gavi introduces flow counters bulk allocation and pool,
   To improve the performance of flow counter acquisition.
3) From Tariq, micro improvements for tx path
4) From Shay, small improvement for XDP TX MPWQE inline flow.
5) Aya provides some cleanups for tx devlink health reporters.
6) Saeed, refactor checksum handling into a single function.
7) Tonghao, allows dropping specific tunnel packets.

----------------------------------------------------------------
Aya Levin (3):
      net/mlx5e: Fix mlx5e_tx_reporter_create return value
      net/mlx5e: Set tx reporter only on successful creation
      net/mlx5e: TX reporter cleanup

Eli Cohen (1):
      net/mlx5: E-Switch, add ingress rate support

Gavi Teitz (2):
      net/mlx5: Add flow counter bulk infrastructure
      net/mlx5: Add flow counter pool

Saeed Mahameed (1):
      net/mlx5e: Rx, checksum handling refactoring

Shay Agroskin (1):
      net/mlx5e: XDP, Close TX MPWQE session when no room for inline packet left

Tariq Toukan (3):
      net/mlx5e: Tx, Strict the room needed for SQ edge NOPs
      net/mlx5e: XDP, Slight enhancement for WQE fetch function
      net/mlx5e: Tx, Soften inline mode VLAN dependencies

Tonghao Zhang (1):
      net/mlx5e: Allow dropping specific tunnel packets

 drivers/net/ethernet/mellanox/mlx5/core/en.h       |   4 +-
 .../net/ethernet/mellanox/mlx5/core/en/reporter.h  |   1 -
 .../ethernet/mellanox/mlx5/core/en/reporter_tx.c   |  17 +-
 drivers/net/ethernet/mellanox/mlx5/core/en/txrx.h  |  24 +-
 drivers/net/ethernet/mellanox/mlx5/core/en/xdp.c   |  36 +--
 drivers/net/ethernet/mellanox/mlx5/core/en/xdp.h   |  66 ++++-
 .../net/ethernet/mellanox/mlx5/core/en_common.c    |  12 -
 drivers/net/ethernet/mellanox/mlx5/core/en_dcbnl.c |   2 +-
 drivers/net/ethernet/mellanox/mlx5/core/en_main.c  |   6 +-
 drivers/net/ethernet/mellanox/mlx5/core/en_rep.c   |  19 ++
 drivers/net/ethernet/mellanox/mlx5/core/en_rep.h   |   1 +
 drivers/net/ethernet/mellanox/mlx5/core/en_rx.c    |  26 +-
 drivers/net/ethernet/mellanox/mlx5/core/en_stats.c |   6 +
 drivers/net/ethernet/mellanox/mlx5/core/en_stats.h |   3 +
 drivers/net/ethernet/mellanox/mlx5/core/en_tc.c    | 103 ++++++-
 drivers/net/ethernet/mellanox/mlx5/core/en_tc.h    |   7 +
 drivers/net/ethernet/mellanox/mlx5/core/en_tx.c    |   7 +-
 drivers/net/ethernet/mellanox/mlx5/core/eswitch.c  |  16 ++
 drivers/net/ethernet/mellanox/mlx5/core/eswitch.h  |   2 +
 .../net/ethernet/mellanox/mlx5/core/fs_counters.c  | 318 +++++++++++++++++++--
 drivers/net/ethernet/mellanox/mlx5/core/vport.c    |   7 +-
 include/linux/mlx5/driver.h                        |  12 +
 22 files changed, 592 insertions(+), 103 deletions(-)

[net-next 01/12] net/mlx5: E-Switch, add ingress rate support

[Saeed Mahameed]

From: Eli Cohen <eli@mellanox.com>

Use the scheduling elements to implement ingress rate limiter on an
eswitch ports ingress traffic. Since the ingress of eswitch port is the
egress of VF port, we control eswitch ingress by controlling VF egress.

Configuration is done using the ports' representor net devices.

Please note that burst size configuration is not supported by devices
ConnectX-5 and earlier generations.

Configuration examples:
tc:
tc filter add dev enp59s0f0_0 root protocol ip matchall action police rate 1mbit burst 20k

ovs:
ovs-vsctl set interface eth0 ingress_policing_rate=1000

Signed-off-by: Eli Cohen <eli@mellanox.com>
Reviewed-by: Paul Blakey <paulb@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
---
 .../net/ethernet/mellanox/mlx5/core/en_rep.c  |  19 ++++
 .../net/ethernet/mellanox/mlx5/core/en_rep.h  |   1 +
 .../net/ethernet/mellanox/mlx5/core/en_tc.c   | 100 ++++++++++++++++++
 .../net/ethernet/mellanox/mlx5/core/en_tc.h   |   7 ++
 .../net/ethernet/mellanox/mlx5/core/eswitch.c |  16 +++
 .../net/ethernet/mellanox/mlx5/core/eswitch.h |   2 +
 6 files changed, 145 insertions(+)

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_rep.c b/drivers/net/ethernet/mellanox/mlx5/core/en_rep.c
index 6edf0aeb1e26..bf6f4835457e 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_rep.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_rep.c
@@ -1156,6 +1156,23 @@ mlx5e_rep_setup_tc_cls_flower(struct mlx5e_priv *priv,
 	}
 }
 
+static
+int mlx5e_rep_setup_tc_cls_matchall(struct mlx5e_priv *priv,
+				    struct tc_cls_matchall_offload *ma)
+{
+	switch (ma->command) {
+	case TC_CLSMATCHALL_REPLACE:
+		return mlx5e_tc_configure_matchall(priv, ma);
+	case TC_CLSMATCHALL_DESTROY:
+		return mlx5e_tc_delete_matchall(priv, ma);
+	case TC_CLSMATCHALL_STATS:
+		mlx5e_tc_stats_matchall(priv, ma);
+		return 0;
+	default:
+		return -EOPNOTSUPP;
+	}
+}
+
 static int mlx5e_rep_setup_tc_cb(enum tc_setup_type type, void *type_data,
 				 void *cb_priv)
 {
@@ -1165,6 +1182,8 @@ static int mlx5e_rep_setup_tc_cb(enum tc_setup_type type, void *type_data,
 	switch (type) {
 	case TC_SETUP_CLSFLOWER:
 		return mlx5e_rep_setup_tc_cls_flower(priv, type_data, flags);
+	case TC_SETUP_CLSMATCHALL:
+		return mlx5e_rep_setup_tc_cls_matchall(priv, type_data);
 	default:
 		return -EOPNOTSUPP;
 	}
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_rep.h b/drivers/net/ethernet/mellanox/mlx5/core/en_rep.h
index 10fafd5fa17b..43eeebe9c8d2 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_rep.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_rep.h
@@ -88,6 +88,7 @@ struct mlx5e_rep_priv {
 	struct mlx5_flow_handle *vport_rx_rule;
 	struct list_head       vport_sqs_list;
 	struct mlx5_rep_uplink_priv uplink_priv; /* valid for uplink rep */
+	struct rtnl_link_stats64 prev_vf_vport_stats;
 	struct devlink_port dl_port;
 };
 
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_tc.c b/drivers/net/ethernet/mellanox/mlx5/core/en_tc.c
index f3ed028d5017..dc5fc3350b65 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_tc.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_tc.c
@@ -3638,6 +3638,106 @@ int mlx5e_stats_flower(struct net_device *dev, struct mlx5e_priv *priv,
 	return err;
 }
 
+static int apply_police_params(struct mlx5e_priv *priv, u32 rate,
+			       struct netlink_ext_ack *extack)
+{
+	struct mlx5e_rep_priv *rpriv = priv->ppriv;
+	struct mlx5_eswitch *esw;
+	u16 vport_num;
+	u32 rate_mbps;
+	int err;
+
+	esw = priv->mdev->priv.eswitch;
+	/* rate is given in bytes/sec.
+	 * First convert to bits/sec and then round to the nearest mbit/secs.
+	 * mbit means million bits.
+	 * Moreover, if rate is non zero we choose to configure to a minimum of
+	 * 1 mbit/sec.
+	 */
+	rate_mbps = rate ? max_t(u32, (rate * 8 + 500000) / 1000000, 1) : 0;
+	vport_num = rpriv->rep->vport;
+
+	err = mlx5_esw_modify_vport_rate(esw, vport_num, rate_mbps);
+	if (err)
+		NL_SET_ERR_MSG_MOD(extack, "failed applying action to hardware");
+
+	return err;
+}
+
+static int scan_tc_matchall_fdb_actions(struct mlx5e_priv *priv,
+					struct flow_action *flow_action,
+					struct netlink_ext_ack *extack)
+{
+	struct mlx5e_rep_priv *rpriv = priv->ppriv;
+	const struct flow_action_entry *act;
+	int err;
+	int i;
+
+	if (!flow_action_has_entries(flow_action)) {
+		NL_SET_ERR_MSG_MOD(extack, "matchall called with no action");
+		return -EINVAL;
+	}
+
+	if (!flow_offload_has_one_action(flow_action)) {
+		NL_SET_ERR_MSG_MOD(extack, "matchall policing support only a single action");
+		return -EOPNOTSUPP;
+	}
+
+	flow_action_for_each(i, act, flow_action) {
+		switch (act->id) {
+		case FLOW_ACTION_POLICE:
+			err = apply_police_params(priv, act->police.rate_bytes_ps, extack);
+			if (err)
+				return err;
+
+			rpriv->prev_vf_vport_stats = priv->stats.vf_vport;
+			break;
+		default:
+			NL_SET_ERR_MSG_MOD(extack, "mlx5 supports only police action for matchall");
+			return -EOPNOTSUPP;
+		}
+	}
+
+	return 0;
+}
+
+int mlx5e_tc_configure_matchall(struct mlx5e_priv *priv,
+				struct tc_cls_matchall_offload *ma)
+{
+	struct netlink_ext_ack *extack = ma->common.extack;
+	int prio = TC_H_MAJ(ma->common.prio) >> 16;
+
+	if (prio != 1) {
+		NL_SET_ERR_MSG_MOD(extack, "only priority 1 is supported");
+		return -EINVAL;
+	}
+
+	return scan_tc_matchall_fdb_actions(priv, &ma->rule->action, extack);
+}
+
+int mlx5e_tc_delete_matchall(struct mlx5e_priv *priv,
+			     struct tc_cls_matchall_offload *ma)
+{
+	struct netlink_ext_ack *extack = ma->common.extack;
+
+	return apply_police_params(priv, 0, extack);
+}
+
+void mlx5e_tc_stats_matchall(struct mlx5e_priv *priv,
+			     struct tc_cls_matchall_offload *ma)
+{
+	struct mlx5e_rep_priv *rpriv = priv->ppriv;
+	struct rtnl_link_stats64 cur_stats;
+	u64 dbytes;
+	u64 dpkts;
+
+	cur_stats = priv->stats.vf_vport;
+	dpkts = cur_stats.rx_packets - rpriv->prev_vf_vport_stats.rx_packets;
+	dbytes = cur_stats.rx_bytes - rpriv->prev_vf_vport_stats.rx_bytes;
+	rpriv->prev_vf_vport_stats = cur_stats;
+	flow_stats_update(&ma->stats, dpkts, dbytes, jiffies);
+}
+
 static void mlx5e_tc_hairpin_update_dead_peer(struct mlx5e_priv *priv,
 					      struct mlx5e_priv *peer_priv)
 {
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_tc.h b/drivers/net/ethernet/mellanox/mlx5/core/en_tc.h
index 1cb66bf76997..20f045e96c92 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_tc.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_tc.h
@@ -63,6 +63,13 @@ int mlx5e_delete_flower(struct net_device *dev, struct mlx5e_priv *priv,
 int mlx5e_stats_flower(struct net_device *dev, struct mlx5e_priv *priv,
 		       struct flow_cls_offload *f, unsigned long flags);
 
+int mlx5e_tc_configure_matchall(struct mlx5e_priv *priv,
+				struct tc_cls_matchall_offload *f);
+int mlx5e_tc_delete_matchall(struct mlx5e_priv *priv,
+			     struct tc_cls_matchall_offload *f);
+void mlx5e_tc_stats_matchall(struct mlx5e_priv *priv,
+			     struct tc_cls_matchall_offload *ma);
+
 struct mlx5e_encap_entry;
 void mlx5e_tc_encap_flows_add(struct mlx5e_priv *priv,
 			      struct mlx5e_encap_entry *e);
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/eswitch.c b/drivers/net/ethernet/mellanox/mlx5/core/eswitch.c
index f4ace5f8e884..5fbebee7254d 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/eswitch.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/eswitch.c
@@ -1585,6 +1585,22 @@ static int esw_vport_qos_config(struct mlx5_eswitch *esw,
 	return 0;
 }
 
+int mlx5_esw_modify_vport_rate(struct mlx5_eswitch *esw, u16 vport_num,
+			       u32 rate_mbps)
+{
+	u32 ctx[MLX5_ST_SZ_DW(scheduling_context)] = {};
+	struct mlx5_vport *vport;
+
+	vport = mlx5_eswitch_get_vport(esw, vport_num);
+	MLX5_SET(scheduling_context, ctx, max_average_bw, rate_mbps);
+
+	return mlx5_modify_scheduling_element_cmd(esw->dev,
+						  SCHEDULING_HIERARCHY_E_SWITCH,
+						  ctx,
+						  vport->qos.esw_tsar_ix,
+						  MODIFY_SCHEDULING_ELEMENT_IN_MODIFY_BITMASK_MAX_AVERAGE_BW);
+}
+
 static void node_guid_gen_from_mac(u64 *node_guid, u8 mac[ETH_ALEN])
 {
 	((u8 *)node_guid)[7] = mac[0];
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/eswitch.h b/drivers/net/ethernet/mellanox/mlx5/core/eswitch.h
index 4a03fdadb47e..804912e38dee 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/eswitch.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/eswitch.h
@@ -261,6 +261,8 @@ void esw_vport_disable_ingress_acl(struct mlx5_eswitch *esw,
 				   struct mlx5_vport *vport);
 void esw_vport_del_ingress_acl_modify_metadata(struct mlx5_eswitch *esw,
 					       struct mlx5_vport *vport);
+int mlx5_esw_modify_vport_rate(struct mlx5_eswitch *esw, u16 vport_num,
+			       u32 rate_mbps);
 
 /* E-Switch API */
 int mlx5_eswitch_init(struct mlx5_core_dev *dev);
-- 
2.21.0

Re: [net-next 01/12] net/mlx5: E-Switch, add ingress rate support

[Alexei Starovoitov]

On Thu, Aug 1, 2019 at 6:30 PM Saeed Mahameed <saeedm@mellanox.com> wrote:
>
> From: Eli Cohen <eli@mellanox.com>
>
> Use the scheduling elements to implement ingress rate limiter on an
> eswitch ports ingress traffic. Since the ingress of eswitch port is the
> egress of VF port, we control eswitch ingress by controlling VF egress.

Looks like the patch is only passing args to firmware which is doing the magic.
Can you please describe what is the algorithm there?
Is it configurable?

Re: [net-next 01/12] net/mlx5: E-Switch, add ingress rate support

[Saeed Mahameed]

On Fri, 2019-08-02 at 10:37 -0700, Alexei Starovoitov wrote:
> On Thu, Aug 1, 2019 at 6:30 PM Saeed Mahameed <saeedm@mellanox.com>
> wrote:
> > From: Eli Cohen <eli@mellanox.com>
> > 
> > Use the scheduling elements to implement ingress rate limiter on an
> > eswitch ports ingress traffic. Since the ingress of eswitch port is
> > the
> > egress of VF port, we control eswitch ingress by controlling VF
> > egress.
> 
> Looks like the patch is only passing args to firmware which is doing
> the magic.
> Can you please describe what is the algorithm there?
> Is it configurable?

Hi Alexei, 

I am not sure how much details you are looking for, but let me share
some of what i know:

From a previous submission for legacy mode sriov vf bw limit, where we 
introduced the FW configuration API and the legacy sriov use case: 
https://patchwork.kernel.org/patch/9404655/

So basically the algorithm is Deficit Weighted Round Robin (DWRR)
between the agents, we can control BW allocation/weight of each agent
(vf vport).

Quoting the commit message from the above link:

"The TSAR implements a Deficit Weighted Round Robin (DWRR) between the
agents. Each agent attached to the TSAR is assigned with a Weight. An
agent is awarded with transmission tokens according to its Weight, and
charged with transmission Tokens according to the amount of data it has
transmitted. Effectively, the Weight (relative to the other agents’
Weight) defines the percentage of the BW an agent will receive,
assuming it has enough data to sustain this BW.

This arbitration scheme is work-preserving, meaning that an agent not
using the entire BW it was allocated, hands over the excess BW, to be
redistributed among the other agents. Each agent will receive
additional BW according to its Weight."


Thanks,
Saeed.

Re: [net-next 01/12] net/mlx5: E-Switch, add ingress rate support

[Alexei Starovoitov]

On Fri, Aug 02, 2019 at 07:22:21PM +0000, Saeed Mahameed wrote:
> On Fri, 2019-08-02 at 10:37 -0700, Alexei Starovoitov wrote:
> > On Thu, Aug 1, 2019 at 6:30 PM Saeed Mahameed <saeedm@mellanox.com>
> > wrote:
> > > From: Eli Cohen <eli@mellanox.com>
> > > 
> > > Use the scheduling elements to implement ingress rate limiter on an
> > > eswitch ports ingress traffic. Since the ingress of eswitch port is
> > > the
> > > egress of VF port, we control eswitch ingress by controlling VF
> > > egress.
> > 
> > Looks like the patch is only passing args to firmware which is doing
> > the magic.
> > Can you please describe what is the algorithm there?
> > Is it configurable?
> 
> Hi Alexei, 
> 
> I am not sure how much details you are looking for, but let me share
> some of what i know:
> 
> From a previous submission for legacy mode sriov vf bw limit, where we 
> introduced the FW configuration API and the legacy sriov use case: 
> https://patchwork.kernel.org/patch/9404655/
> 
> So basically the algorithm is Deficit Weighted Round Robin (DWRR)
> between the agents, we can control BW allocation/weight of each agent
> (vf vport).

commit log of this patch says nothing about DWRR.
It is also not using any of the api that were provided by that
earlier patch.
what is going on?

Re: [net-next 01/12] net/mlx5: E-Switch, add ingress rate support

[Saeed Mahameed]

On Fri, 2019-08-02 at 12:44 -0700, Alexei Starovoitov wrote:
> On Fri, Aug 02, 2019 at 07:22:21PM +0000, Saeed Mahameed wrote:
> > On Fri, 2019-08-02 at 10:37 -0700, Alexei Starovoitov wrote:
> > > On Thu, Aug 1, 2019 at 6:30 PM Saeed Mahameed <
> > > saeedm@mellanox.com>
> > > wrote:
> > > > From: Eli Cohen <eli@mellanox.com>
> > > > 
> > > > Use the scheduling elements to implement ingress rate limiter
> > > > on an
> > > > eswitch ports ingress traffic. Since the ingress of eswitch
> > > > port is
> > > > the
> > > > egress of VF port, we control eswitch ingress by controlling VF
> > > > egress.
> > > 
> > > Looks like the patch is only passing args to firmware which is
> > > doing
> > > the magic.
> > > Can you please describe what is the algorithm there?
> > > Is it configurable?
> > 
> > Hi Alexei, 
> > 
> > I am not sure how much details you are looking for, but let me
> > share
> > some of what i know:
> > 
> > From a previous submission for legacy mode sriov vf bw limit, where
> > we 
> > introduced the FW configuration API and the legacy sriov use case: 
> > https://patchwork.kernel.org/patch/9404655/
> > 
> > So basically the algorithm is Deficit Weighted Round Robin (DWRR)
> > between the agents, we can control BW allocation/weight of each
> > agent
> > (vf vport).
> 
> commit log of this patch says nothing about DWRR.

it says it uses the "scheduling elements ingress rate limiter" which
automatically translates to DWRR, this is basic knowledge for mlx5
developers, and it is clear in the commit message when the FW API was
introduced.

> It is also not using any of the api that were provided by that
> earlier patch.

It is using the api: 
mlx5e_tc_configure_matchall => apply_police_params =>
mlx5_esw_modify_vport_rate => mlx5_modify_scheduling_element_cmd()..

> what is going on?
> 

can you please clarify what is bothering you ? we can fix it if
necessary.

Thanks,
Saeed.

[net-next 02/12] net/mlx5: Add flow counter bulk infrastructure

[Saeed Mahameed]

From: Gavi Teitz <gavi@mellanox.com>

Add infrastructure to track bulks of flow counters, providing
the means to allocate and deallocate bulks, and to acquire and
release individual counters from the bulks.

Signed-off-by: Gavi Teitz <gavi@mellanox.com>
Reviewed-by: Vlad Buslov <vladbu@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
---
 .../ethernet/mellanox/mlx5/core/fs_counters.c | 105 ++++++++++++++++++
 1 file changed, 105 insertions(+)

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/fs_counters.c b/drivers/net/ethernet/mellanox/mlx5/core/fs_counters.c
index 067a4b56498b..3e734e62a6cd 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/fs_counters.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/fs_counters.c
@@ -58,6 +58,7 @@ struct mlx5_fc {
 	u64 lastpackets;
 	u64 lastbytes;
 
+	struct mlx5_fc_bulk *bulk;
 	u32 id;
 	bool aging;
 
@@ -412,3 +413,107 @@ void mlx5_fc_update_sampling_interval(struct mlx5_core_dev *dev,
 	fc_stats->sampling_interval = min_t(unsigned long, interval,
 					    fc_stats->sampling_interval);
 }
+
+/* Flow counter bluks */
+
+struct mlx5_fc_bulk {
+	u32 base_id;
+	int bulk_len;
+	unsigned long *bitmask;
+	struct mlx5_fc fcs[0];
+};
+
+static void
+mlx5_fc_init(struct mlx5_fc *counter, struct mlx5_fc_bulk *bulk, u32 id)
+{
+	counter->bulk = bulk;
+	counter->id = id;
+}
+
+static int mlx5_fc_bulk_get_free_fcs_amount(struct mlx5_fc_bulk *bulk)
+{
+	return bitmap_weight(bulk->bitmask, bulk->bulk_len);
+}
+
+static struct mlx5_fc_bulk __attribute__((unused))
+*mlx5_fc_bulk_create(struct mlx5_core_dev *dev)
+{
+	enum mlx5_fc_bulk_alloc_bitmask alloc_bitmask;
+	struct mlx5_fc_bulk *bulk;
+	int err = -ENOMEM;
+	int bulk_len;
+	u32 base_id;
+	int i;
+
+	alloc_bitmask = MLX5_CAP_GEN(dev, flow_counter_bulk_alloc);
+	bulk_len = alloc_bitmask > 0 ? MLX5_FC_BULK_NUM_FCS(alloc_bitmask) : 1;
+
+	bulk = kzalloc(sizeof(*bulk) + bulk_len * sizeof(struct mlx5_fc),
+		       GFP_KERNEL);
+	if (!bulk)
+		goto err_alloc_bulk;
+
+	bulk->bitmask = kcalloc(BITS_TO_LONGS(bulk_len), sizeof(unsigned long),
+				GFP_KERNEL);
+	if (!bulk->bitmask)
+		goto err_alloc_bitmask;
+
+	err = mlx5_cmd_fc_bulk_alloc(dev, alloc_bitmask, &base_id);
+	if (err)
+		goto err_mlx5_cmd_bulk_alloc;
+
+	bulk->base_id = base_id;
+	bulk->bulk_len = bulk_len;
+	for (i = 0; i < bulk_len; i++) {
+		mlx5_fc_init(&bulk->fcs[i], bulk, base_id + i);
+		set_bit(i, bulk->bitmask);
+	}
+
+	return bulk;
+
+err_mlx5_cmd_bulk_alloc:
+	kfree(bulk->bitmask);
+err_alloc_bitmask:
+	kfree(bulk);
+err_alloc_bulk:
+	return ERR_PTR(err);
+}
+
+static int __attribute__((unused))
+mlx5_fc_bulk_destroy(struct mlx5_core_dev *dev, struct mlx5_fc_bulk *bulk)
+{
+	if (mlx5_fc_bulk_get_free_fcs_amount(bulk) < bulk->bulk_len) {
+		mlx5_core_err(dev, "Freeing bulk before all counters were released\n");
+		return -EBUSY;
+	}
+
+	mlx5_cmd_fc_free(dev, bulk->base_id);
+	kfree(bulk->bitmask);
+	kfree(bulk);
+
+	return 0;
+}
+
+static struct mlx5_fc __attribute__((unused))
+*mlx5_fc_bulk_acquire_fc(struct mlx5_fc_bulk *bulk)
+{
+	int free_fc_index = find_first_bit(bulk->bitmask, bulk->bulk_len);
+
+	if (free_fc_index >= bulk->bulk_len)
+		return ERR_PTR(-ENOSPC);
+
+	clear_bit(free_fc_index, bulk->bitmask);
+	return &bulk->fcs[free_fc_index];
+}
+
+static int __attribute__((unused))
+mlx5_fc_bulk_release_fc(struct mlx5_fc_bulk *bulk, struct mlx5_fc *fc)
+{
+	int fc_index = fc->id - bulk->base_id;
+
+	if (test_bit(fc_index, bulk->bitmask))
+		return -EINVAL;
+
+	set_bit(fc_index, bulk->bitmask);
+	return 0;
+}
-- 
2.21.0

[net-next 03/12] net/mlx5: Add flow counter pool

[Saeed Mahameed]

From: Gavi Teitz <gavi@mellanox.com>

Add a pool of flow counters, based on flow counter bulks, removing the
need to allocate a new counter via a costly FW command during the flow
creation process. The time it takes to acquire/release a flow counter
is cut from ~50 [us] to ~50 [ns].

The pool is part of the mlx5 driver instance, and provides flow
counters for aging flows. mlx5_fc_create() was modified to provide
counters for aging flows from the pool by default, and
mlx5_destroy_fc() was modified to release counters back to the pool
for later reuse. If bulk allocation is not supported or fails, and for
non-aging flows, the fallback behavior is to allocate and free
individual counters.

The pool is comprised of three lists of flow counter bulks, one of
fully used bulks, one of partially used bulks, and one of unused
bulks. Counters are provided from the partially used bulks first, to
help limit bulk fragmentation.

The pool maintains a threshold, and strives to maintain the amount of
available counters below it. The pool is increased in size when a
counter acquisition request is made and there are no available
counters, and it is decreased in size when the last counter in a bulk
is released and there are more available counters than the threshold.
All pool size changes are done in the context of the
acquiring/releasing process.

The value of the threshold is directly correlated to the amount of
used counters the pool is providing, while constrained by a hard
maximum, and is recalculated every time a bulk is allocated/freed.
This ensures that the pool only consumes large amounts of memory for
available counters if the pool is being used heavily. When fully
populated and at the hard maximum, the buffer of available counters
consumes ~40 [MB].

Signed-off-by: Gavi Teitz <gavi@mellanox.com>
Reviewed-by: Vlad Buslov <vladbu@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
---
 .../ethernet/mellanox/mlx5/core/fs_counters.c | 231 ++++++++++++++++--
 include/linux/mlx5/driver.h                   |  12 +
 2 files changed, 218 insertions(+), 25 deletions(-)

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/fs_counters.c b/drivers/net/ethernet/mellanox/mlx5/core/fs_counters.c
index 3e734e62a6cd..51f1736c455d 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/fs_counters.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/fs_counters.c
@@ -40,6 +40,8 @@
 #define MLX5_FC_STATS_PERIOD msecs_to_jiffies(1000)
 /* Max number of counters to query in bulk read is 32K */
 #define MLX5_SW_MAX_COUNTERS_BULK BIT(15)
+#define MLX5_FC_POOL_MAX_THRESHOLD BIT(18)
+#define MLX5_FC_POOL_USED_BUFF_RATIO 10
 
 struct mlx5_fc_cache {
 	u64 packets;
@@ -65,6 +67,11 @@ struct mlx5_fc {
 	struct mlx5_fc_cache cache ____cacheline_aligned_in_smp;
 };
 
+static void mlx5_fc_pool_init(struct mlx5_fc_pool *fc_pool, struct mlx5_core_dev *dev);
+static void mlx5_fc_pool_cleanup(struct mlx5_fc_pool *fc_pool);
+static struct mlx5_fc *mlx5_fc_pool_acquire_counter(struct mlx5_fc_pool *fc_pool);
+static void mlx5_fc_pool_release_counter(struct mlx5_fc_pool *fc_pool, struct mlx5_fc *fc);
+
 /* locking scheme:
  *
  * It is the responsibility of the user to prevent concurrent calls or bad
@@ -202,13 +209,22 @@ static void mlx5_fc_stats_query_counter_range(struct mlx5_core_dev *dev,
 	}
 }
 
-static void mlx5_free_fc(struct mlx5_core_dev *dev,
-			 struct mlx5_fc *counter)
+static void mlx5_fc_free(struct mlx5_core_dev *dev, struct mlx5_fc *counter)
 {
 	mlx5_cmd_fc_free(dev, counter->id);
 	kfree(counter);
 }
 
+static void mlx5_fc_release(struct mlx5_core_dev *dev, struct mlx5_fc *counter)
+{
+	struct mlx5_fc_stats *fc_stats = &dev->priv.fc_stats;
+
+	if (counter->bulk)
+		mlx5_fc_pool_release_counter(&fc_stats->fc_pool, counter);
+	else
+		mlx5_fc_free(dev, counter);
+}
+
 static void mlx5_fc_stats_work(struct work_struct *work)
 {
 	struct mlx5_core_dev *dev = container_of(work, struct mlx5_core_dev,
@@ -232,7 +248,7 @@ static void mlx5_fc_stats_work(struct work_struct *work)
 	llist_for_each_entry_safe(counter, tmp, dellist, dellist) {
 		mlx5_fc_stats_remove(dev, counter);
 
-		mlx5_free_fc(dev, counter);
+		mlx5_fc_release(dev, counter);
 	}
 
 	if (time_before(now, fc_stats->next_query) ||
@@ -248,26 +264,56 @@ static void mlx5_fc_stats_work(struct work_struct *work)
 	fc_stats->next_query = now + fc_stats->sampling_interval;
 }
 
-struct mlx5_fc *mlx5_fc_create(struct mlx5_core_dev *dev, bool aging)
+static struct mlx5_fc *mlx5_fc_single_alloc(struct mlx5_core_dev *dev)
 {
-	struct mlx5_fc_stats *fc_stats = &dev->priv.fc_stats;
 	struct mlx5_fc *counter;
 	int err;
 
 	counter = kzalloc(sizeof(*counter), GFP_KERNEL);
 	if (!counter)
 		return ERR_PTR(-ENOMEM);
-	INIT_LIST_HEAD(&counter->list);
 
 	err = mlx5_cmd_fc_alloc(dev, &counter->id);
-	if (err)
-		goto err_out;
+	if (err) {
+		kfree(counter);
+		return ERR_PTR(err);
+	}
+
+	return counter;
+}
+
+static struct mlx5_fc *mlx5_fc_acquire(struct mlx5_core_dev *dev, bool aging)
+{
+	struct mlx5_fc_stats *fc_stats = &dev->priv.fc_stats;
+	struct mlx5_fc *counter;
+
+	if (aging && MLX5_CAP_GEN(dev, flow_counter_bulk_alloc) != 0) {
+		counter = mlx5_fc_pool_acquire_counter(&fc_stats->fc_pool);
+		if (!IS_ERR(counter))
+			return counter;
+	}
+
+	return mlx5_fc_single_alloc(dev);
+}
+
+struct mlx5_fc *mlx5_fc_create(struct mlx5_core_dev *dev, bool aging)
+{
+	struct mlx5_fc *counter = mlx5_fc_acquire(dev, aging);
+	struct mlx5_fc_stats *fc_stats = &dev->priv.fc_stats;
+	int err;
+
+	if (IS_ERR(counter))
+		return counter;
+
+	INIT_LIST_HEAD(&counter->list);
+	counter->aging = aging;
 
 	if (aging) {
 		u32 id = counter->id;
 
 		counter->cache.lastuse = jiffies;
-		counter->aging = true;
+		counter->lastbytes = counter->cache.bytes;
+		counter->lastpackets = counter->cache.packets;
 
 		idr_preload(GFP_KERNEL);
 		spin_lock(&fc_stats->counters_idr_lock);
@@ -288,10 +334,7 @@ struct mlx5_fc *mlx5_fc_create(struct mlx5_core_dev *dev, bool aging)
 	return counter;
 
 err_out_alloc:
-	mlx5_cmd_fc_free(dev, counter->id);
-err_out:
-	kfree(counter);
-
+	mlx5_fc_release(dev, counter);
 	return ERR_PTR(err);
 }
 EXPORT_SYMBOL(mlx5_fc_create);
@@ -315,7 +358,7 @@ void mlx5_fc_destroy(struct mlx5_core_dev *dev, struct mlx5_fc *counter)
 		return;
 	}
 
-	mlx5_free_fc(dev, counter);
+	mlx5_fc_release(dev, counter);
 }
 EXPORT_SYMBOL(mlx5_fc_destroy);
 
@@ -344,6 +387,7 @@ int mlx5_init_fc_stats(struct mlx5_core_dev *dev)
 	fc_stats->sampling_interval = MLX5_FC_STATS_PERIOD;
 	INIT_DELAYED_WORK(&fc_stats->work, mlx5_fc_stats_work);
 
+	mlx5_fc_pool_init(&fc_stats->fc_pool, dev);
 	return 0;
 
 err_wq_create:
@@ -358,6 +402,7 @@ void mlx5_cleanup_fc_stats(struct mlx5_core_dev *dev)
 	struct mlx5_fc *counter;
 	struct mlx5_fc *tmp;
 
+	mlx5_fc_pool_cleanup(&fc_stats->fc_pool);
 	cancel_delayed_work_sync(&dev->priv.fc_stats.work);
 	destroy_workqueue(dev->priv.fc_stats.wq);
 	dev->priv.fc_stats.wq = NULL;
@@ -368,10 +413,10 @@ void mlx5_cleanup_fc_stats(struct mlx5_core_dev *dev)
 
 	tmplist = llist_del_all(&fc_stats->addlist);
 	llist_for_each_entry_safe(counter, tmp, tmplist, addlist)
-		mlx5_free_fc(dev, counter);
+		mlx5_fc_release(dev, counter);
 
 	list_for_each_entry_safe(counter, tmp, &fc_stats->counters, list)
-		mlx5_free_fc(dev, counter);
+		mlx5_fc_release(dev, counter);
 }
 
 int mlx5_fc_query(struct mlx5_core_dev *dev, struct mlx5_fc *counter,
@@ -417,14 +462,15 @@ void mlx5_fc_update_sampling_interval(struct mlx5_core_dev *dev,
 /* Flow counter bluks */
 
 struct mlx5_fc_bulk {
+	struct list_head pool_list;
 	u32 base_id;
 	int bulk_len;
 	unsigned long *bitmask;
 	struct mlx5_fc fcs[0];
 };
 
-static void
-mlx5_fc_init(struct mlx5_fc *counter, struct mlx5_fc_bulk *bulk, u32 id)
+static void mlx5_fc_init(struct mlx5_fc *counter, struct mlx5_fc_bulk *bulk,
+			 u32 id)
 {
 	counter->bulk = bulk;
 	counter->id = id;
@@ -435,8 +481,7 @@ static int mlx5_fc_bulk_get_free_fcs_amount(struct mlx5_fc_bulk *bulk)
 	return bitmap_weight(bulk->bitmask, bulk->bulk_len);
 }
 
-static struct mlx5_fc_bulk __attribute__((unused))
-*mlx5_fc_bulk_create(struct mlx5_core_dev *dev)
+static struct mlx5_fc_bulk *mlx5_fc_bulk_create(struct mlx5_core_dev *dev)
 {
 	enum mlx5_fc_bulk_alloc_bitmask alloc_bitmask;
 	struct mlx5_fc_bulk *bulk;
@@ -479,7 +524,7 @@ static struct mlx5_fc_bulk __attribute__((unused))
 	return ERR_PTR(err);
 }
 
-static int __attribute__((unused))
+static int
 mlx5_fc_bulk_destroy(struct mlx5_core_dev *dev, struct mlx5_fc_bulk *bulk)
 {
 	if (mlx5_fc_bulk_get_free_fcs_amount(bulk) < bulk->bulk_len) {
@@ -494,8 +539,7 @@ mlx5_fc_bulk_destroy(struct mlx5_core_dev *dev, struct mlx5_fc_bulk *bulk)
 	return 0;
 }
 
-static struct mlx5_fc __attribute__((unused))
-*mlx5_fc_bulk_acquire_fc(struct mlx5_fc_bulk *bulk)
+static struct mlx5_fc *mlx5_fc_bulk_acquire_fc(struct mlx5_fc_bulk *bulk)
 {
 	int free_fc_index = find_first_bit(bulk->bitmask, bulk->bulk_len);
 
@@ -506,8 +550,7 @@ static struct mlx5_fc __attribute__((unused))
 	return &bulk->fcs[free_fc_index];
 }
 
-static int __attribute__((unused))
-mlx5_fc_bulk_release_fc(struct mlx5_fc_bulk *bulk, struct mlx5_fc *fc)
+static int mlx5_fc_bulk_release_fc(struct mlx5_fc_bulk *bulk, struct mlx5_fc *fc)
 {
 	int fc_index = fc->id - bulk->base_id;
 
@@ -517,3 +560,141 @@ mlx5_fc_bulk_release_fc(struct mlx5_fc_bulk *bulk, struct mlx5_fc *fc)
 	set_bit(fc_index, bulk->bitmask);
 	return 0;
 }
+
+/* Flow counters pool API */
+
+static void mlx5_fc_pool_init(struct mlx5_fc_pool *fc_pool, struct mlx5_core_dev *dev)
+{
+	fc_pool->dev = dev;
+	mutex_init(&fc_pool->pool_lock);
+	INIT_LIST_HEAD(&fc_pool->fully_used);
+	INIT_LIST_HEAD(&fc_pool->partially_used);
+	INIT_LIST_HEAD(&fc_pool->unused);
+	fc_pool->available_fcs = 0;
+	fc_pool->used_fcs = 0;
+	fc_pool->threshold = 0;
+}
+
+static void mlx5_fc_pool_cleanup(struct mlx5_fc_pool *fc_pool)
+{
+	struct mlx5_core_dev *dev = fc_pool->dev;
+	struct mlx5_fc_bulk *bulk;
+	struct mlx5_fc_bulk *tmp;
+
+	list_for_each_entry_safe(bulk, tmp, &fc_pool->fully_used, pool_list)
+		mlx5_fc_bulk_destroy(dev, bulk);
+	list_for_each_entry_safe(bulk, tmp, &fc_pool->partially_used, pool_list)
+		mlx5_fc_bulk_destroy(dev, bulk);
+	list_for_each_entry_safe(bulk, tmp, &fc_pool->unused, pool_list)
+		mlx5_fc_bulk_destroy(dev, bulk);
+}
+
+static void mlx5_fc_pool_update_threshold(struct mlx5_fc_pool *fc_pool)
+{
+	fc_pool->threshold = min_t(int, MLX5_FC_POOL_MAX_THRESHOLD,
+				   fc_pool->used_fcs / MLX5_FC_POOL_USED_BUFF_RATIO);
+}
+
+static struct mlx5_fc_bulk *
+mlx5_fc_pool_alloc_new_bulk(struct mlx5_fc_pool *fc_pool)
+{
+	struct mlx5_core_dev *dev = fc_pool->dev;
+	struct mlx5_fc_bulk *new_bulk;
+
+	new_bulk = mlx5_fc_bulk_create(dev);
+	if (!IS_ERR(new_bulk))
+		fc_pool->available_fcs += new_bulk->bulk_len;
+	mlx5_fc_pool_update_threshold(fc_pool);
+	return new_bulk;
+}
+
+static void
+mlx5_fc_pool_free_bulk(struct mlx5_fc_pool *fc_pool, struct mlx5_fc_bulk *bulk)
+{
+	struct mlx5_core_dev *dev = fc_pool->dev;
+
+	fc_pool->available_fcs -= bulk->bulk_len;
+	mlx5_fc_bulk_destroy(dev, bulk);
+	mlx5_fc_pool_update_threshold(fc_pool);
+}
+
+static struct mlx5_fc *
+mlx5_fc_pool_acquire_from_list(struct list_head *src_list,
+			       struct list_head *next_list,
+			       bool move_non_full_bulk)
+{
+	struct mlx5_fc_bulk *bulk;
+	struct mlx5_fc *fc;
+
+	if (list_empty(src_list))
+		return ERR_PTR(-ENODATA);
+
+	bulk = list_first_entry(src_list, struct mlx5_fc_bulk, pool_list);
+	fc = mlx5_fc_bulk_acquire_fc(bulk);
+	if (move_non_full_bulk || mlx5_fc_bulk_get_free_fcs_amount(bulk) == 0)
+		list_move(&bulk->pool_list, next_list);
+	return fc;
+}
+
+static struct mlx5_fc *
+mlx5_fc_pool_acquire_counter(struct mlx5_fc_pool *fc_pool)
+{
+	struct mlx5_fc_bulk *new_bulk;
+	struct mlx5_fc *fc;
+
+	mutex_lock(&fc_pool->pool_lock);
+
+	fc = mlx5_fc_pool_acquire_from_list(&fc_pool->partially_used,
+					    &fc_pool->fully_used, false);
+	if (IS_ERR(fc))
+		fc = mlx5_fc_pool_acquire_from_list(&fc_pool->unused,
+						    &fc_pool->partially_used,
+						    true);
+	if (IS_ERR(fc)) {
+		new_bulk = mlx5_fc_pool_alloc_new_bulk(fc_pool);
+		if (IS_ERR(new_bulk)) {
+			fc = ERR_CAST(new_bulk);
+			goto out;
+		}
+		fc = mlx5_fc_bulk_acquire_fc(new_bulk);
+		list_add(&new_bulk->pool_list, &fc_pool->partially_used);
+	}
+	fc_pool->available_fcs--;
+	fc_pool->used_fcs++;
+
+out:
+	mutex_unlock(&fc_pool->pool_lock);
+	return fc;
+}
+
+static void
+mlx5_fc_pool_release_counter(struct mlx5_fc_pool *fc_pool, struct mlx5_fc *fc)
+{
+	struct mlx5_core_dev *dev = fc_pool->dev;
+	struct mlx5_fc_bulk *bulk = fc->bulk;
+	int bulk_free_fcs_amount;
+
+	mutex_lock(&fc_pool->pool_lock);
+
+	if (mlx5_fc_bulk_release_fc(bulk, fc)) {
+		mlx5_core_warn(dev, "Attempted to release a counter which is not acquired\n");
+		goto unlock;
+	}
+
+	fc_pool->available_fcs++;
+	fc_pool->used_fcs--;
+
+	bulk_free_fcs_amount = mlx5_fc_bulk_get_free_fcs_amount(bulk);
+	if (bulk_free_fcs_amount == 1)
+		list_move_tail(&bulk->pool_list, &fc_pool->partially_used);
+	if (bulk_free_fcs_amount == bulk->bulk_len) {
+		list_del(&bulk->pool_list);
+		if (fc_pool->available_fcs > fc_pool->threshold)
+			mlx5_fc_pool_free_bulk(fc_pool, bulk);
+		else
+			list_add(&bulk->pool_list, &fc_pool->unused);
+	}
+
+unlock:
+	mutex_unlock(&fc_pool->pool_lock);
+}
diff --git a/include/linux/mlx5/driver.h b/include/linux/mlx5/driver.h
index 267b2bc0ca4a..d8f348ef9c33 100644
--- a/include/linux/mlx5/driver.h
+++ b/include/linux/mlx5/driver.h
@@ -477,6 +477,17 @@ struct mlx5_core_sriov {
 	u16			max_vfs;
 };
 
+struct mlx5_fc_pool {
+	struct mlx5_core_dev *dev;
+	struct mutex pool_lock; /* protects pool lists */
+	struct list_head fully_used;
+	struct list_head partially_used;
+	struct list_head unused;
+	int available_fcs;
+	int used_fcs;
+	int threshold;
+};
+
 struct mlx5_fc_stats {
 	spinlock_t counters_idr_lock; /* protects counters_idr */
 	struct idr counters_idr;
@@ -489,6 +500,7 @@ struct mlx5_fc_stats {
 	unsigned long next_query;
 	unsigned long sampling_interval; /* jiffies */
 	u32 *bulk_query_out;
+	struct mlx5_fc_pool fc_pool;
 };
 
 struct mlx5_events;
-- 
2.21.0

[net-next 04/12] net/mlx5e: Tx, Strict the room needed for SQ edge NOPs

[Saeed Mahameed]

From: Tariq Toukan <tariqt@mellanox.com>

We use NOPs to populate the WQ fragment edge if the WQE does not fit
in frag, to avoid WQEs crossing a page boundary (or wrap-around the WQ).

The upper bound on the needed number of NOPs is one WQEBB less than
the largest possible WQE, for otherwise the WQE would certainly fit.

Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
---
 drivers/net/ethernet/mellanox/mlx5/core/en/txrx.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/txrx.h b/drivers/net/ethernet/mellanox/mlx5/core/en/txrx.h
index ddfe19adb3d9..7da22b413a48 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en/txrx.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en/txrx.h
@@ -6,7 +6,7 @@
 
 #include "en.h"
 
-#define MLX5E_SQ_NOPS_ROOM  MLX5_SEND_WQE_MAX_WQEBBS
+#define MLX5E_SQ_NOPS_ROOM (MLX5_SEND_WQE_MAX_WQEBBS - 1)
 #define MLX5E_SQ_STOP_ROOM (MLX5_SEND_WQE_MAX_WQEBBS +\
 			    MLX5E_SQ_NOPS_ROOM)
 
-- 
2.21.0

[net-next 05/12] net/mlx5e: XDP, Close TX MPWQE session when no room for inline packet left

[Saeed Mahameed]

From: Shay Agroskin <shayag@mellanox.com>

In MPWQE mode, when transmitting packets with XDP, a packet that is smaller
than a certain size (set to 256 bytes) would be sent inline within its WQE
TX descriptor (mem-copied), in case the hardware tx queue is congested
beyond a pre-defined water-mark.

If a MPWQE cannot contain an additional inline packet, we close this
MPWQE session, and send the packet inlined within the next MPWQE.
To save some MPWQE session close+open operations, we don't open MPWQE
sessions that are contiguously smaller than certain size (set to the
HW MPWQE maximum size). If there isn't enough contiguous room in the
send queue, we fill it with NOPs and wrap the send queue index around.

This way, qualified packets are always sent inline.

Perf tests:
Tested packet rate for UDP 64Byte multi-stream
over two dual port ConnectX-5 100Gbps NICs.
CPU: Intel(R) Xeon(R) CPU E5-2680 v3 @ 2.50GHz

XDP_TX:

With 24 channels:
| ------ | bounced packets | inlined packets | inline ratio |
| before | 113.6Mpps       | 96.3Mpps        | 84%          |
| after  |   115Mpps       | 99.5Mpps        | 86%          |

With one channel:

| ------ | bounced packets | inlined packets | inline ratio |
| before | 6.7Mpps         | 0pps            | 0%           |
| after  | 6.8Mpps         | 0pps            | 0%           |

As we can see, there is improvement in both inline ratio and overall
packet rate for 24 channels. Also, we see no degradation for the
one-channel case.

Signed-off-by: Shay Agroskin <shayag@mellanox.com>
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
---
 drivers/net/ethernet/mellanox/mlx5/core/en.h  |  2 -
 .../net/ethernet/mellanox/mlx5/core/en/xdp.c  | 32 ++++-------
 .../net/ethernet/mellanox/mlx5/core/en/xdp.h  | 53 +++++++++++++++----
 .../ethernet/mellanox/mlx5/core/en_stats.c    |  6 +++
 .../ethernet/mellanox/mlx5/core/en_stats.h    |  3 ++
 5 files changed, 63 insertions(+), 33 deletions(-)

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en.h b/drivers/net/ethernet/mellanox/mlx5/core/en.h
index 79d93d6c7d7a..745bcc25c6f8 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en.h
@@ -483,8 +483,6 @@ struct mlx5e_xdp_mpwqe {
 	struct mlx5e_tx_wqe *wqe;
 	u8                   ds_count;
 	u8                   pkt_count;
-	u8                   max_ds_count;
-	u8                   complete;
 	u8                   inline_on;
 };
 
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.c b/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.c
index b0b982cf69bb..8cb98326531f 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.c
@@ -179,34 +179,22 @@ static void mlx5e_xdp_mpwqe_session_start(struct mlx5e_xdpsq *sq)
 	struct mlx5e_xdp_mpwqe *session = &sq->mpwqe;
 	struct mlx5e_xdpsq_stats *stats = sq->stats;
 	struct mlx5_wq_cyc *wq = &sq->wq;
-	u8  wqebbs;
-	u16 pi;
+	u16 pi, contig_wqebbs;
+
+	pi = mlx5_wq_cyc_ctr2ix(wq, sq->pc);
+	contig_wqebbs = mlx5_wq_cyc_get_contig_wqebbs(wq, pi);
+
+	if (unlikely(contig_wqebbs < MLX5_SEND_WQE_MAX_WQEBBS))
+		mlx5e_fill_xdpsq_frag_edge(sq, wq, pi, contig_wqebbs);
 
 	mlx5e_xdpsq_fetch_wqe(sq, &session->wqe);
 
 	prefetchw(session->wqe->data);
 	session->ds_count  = MLX5E_XDP_TX_EMPTY_DS_COUNT;
 	session->pkt_count = 0;
-	session->complete  = 0;
 
 	pi = mlx5_wq_cyc_ctr2ix(wq, sq->pc);
 
-/* The mult of MLX5_SEND_WQE_MAX_WQEBBS * MLX5_SEND_WQEBB_NUM_DS
- * (16 * 4 == 64) does not fit in the 6-bit DS field of Ctrl Segment.
- * We use a bound lower that MLX5_SEND_WQE_MAX_WQEBBS to let a
- * full-session WQE be cache-aligned.
- */
-#if L1_CACHE_BYTES < 128
-#define MLX5E_XDP_MPW_MAX_WQEBBS (MLX5_SEND_WQE_MAX_WQEBBS - 1)
-#else
-#define MLX5E_XDP_MPW_MAX_WQEBBS (MLX5_SEND_WQE_MAX_WQEBBS - 2)
-#endif
-
-	wqebbs = min_t(u16, mlx5_wq_cyc_get_contig_wqebbs(wq, pi),
-		       MLX5E_XDP_MPW_MAX_WQEBBS);
-
-	session->max_ds_count = MLX5_SEND_WQEBB_NUM_DS * wqebbs;
-
 	mlx5e_xdp_update_inline_state(sq);
 
 	stats->mpwqe++;
@@ -244,7 +232,7 @@ static int mlx5e_xmit_xdp_frame_check_mpwqe(struct mlx5e_xdpsq *sq)
 {
 	if (unlikely(!sq->mpwqe.wqe)) {
 		if (unlikely(!mlx5e_wqc_has_room_for(&sq->wq, sq->cc, sq->pc,
-						     MLX5_SEND_WQE_MAX_WQEBBS))) {
+						     MLX5E_XDPSQ_STOP_ROOM))) {
 			/* SQ is full, ring doorbell */
 			mlx5e_xmit_xdp_doorbell(sq);
 			sq->stats->full++;
@@ -285,8 +273,8 @@ static bool mlx5e_xmit_xdp_frame_mpwqe(struct mlx5e_xdpsq *sq,
 
 	mlx5e_xdp_mpwqe_add_dseg(sq, xdptxd, stats);
 
-	if (unlikely(session->complete ||
-		     session->ds_count == session->max_ds_count))
+	if (unlikely(mlx5e_xdp_no_room_for_inline_pkt(session) ||
+		     session->ds_count == MLX5E_XDP_MPW_MAX_NUM_DS))
 		mlx5e_xdp_mpwqe_complete(sq);
 
 	mlx5e_xdpi_fifo_push(&sq->db.xdpi_fifo, xdpi);
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.h b/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.h
index b90923932668..e0ed7710f5f1 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.h
@@ -40,6 +40,26 @@
 	(sizeof(struct mlx5e_tx_wqe) / MLX5_SEND_WQE_DS)
 #define MLX5E_XDP_TX_DS_COUNT (MLX5E_XDP_TX_EMPTY_DS_COUNT + 1 /* SG DS */)
 
+#define MLX5E_XDPSQ_STOP_ROOM (MLX5E_SQ_STOP_ROOM)
+
+#define MLX5E_XDP_INLINE_WQE_SZ_THRSD (256 - sizeof(struct mlx5_wqe_inline_seg))
+#define MLX5E_XDP_INLINE_WQE_MAX_DS_CNT \
+	DIV_ROUND_UP(MLX5E_XDP_INLINE_WQE_SZ_THRSD, MLX5_SEND_WQE_DS)
+
+/* The mult of MLX5_SEND_WQE_MAX_WQEBBS * MLX5_SEND_WQEBB_NUM_DS
+ * (16 * 4 == 64) does not fit in the 6-bit DS field of Ctrl Segment.
+ * We use a bound lower that MLX5_SEND_WQE_MAX_WQEBBS to let a
+ * full-session WQE be cache-aligned.
+ */
+#if L1_CACHE_BYTES < 128
+#define MLX5E_XDP_MPW_MAX_WQEBBS (MLX5_SEND_WQE_MAX_WQEBBS - 1)
+#else
+#define MLX5E_XDP_MPW_MAX_WQEBBS (MLX5_SEND_WQE_MAX_WQEBBS - 2)
+#endif
+
+#define MLX5E_XDP_MPW_MAX_NUM_DS \
+	(MLX5E_XDP_MPW_MAX_WQEBBS * MLX5_SEND_WQEBB_NUM_DS)
+
 struct mlx5e_xsk_param;
 int mlx5e_xdp_max_mtu(struct mlx5e_params *params, struct mlx5e_xsk_param *xsk);
 bool mlx5e_xdp_handle(struct mlx5e_rq *rq, struct mlx5e_dma_info *di,
@@ -114,6 +134,30 @@ static inline void mlx5e_xdp_update_inline_state(struct mlx5e_xdpsq *sq)
 		session->inline_on = 1;
 }
 
+static inline bool
+mlx5e_xdp_no_room_for_inline_pkt(struct mlx5e_xdp_mpwqe *session)
+{
+	return session->inline_on &&
+	       session->ds_count + MLX5E_XDP_INLINE_WQE_MAX_DS_CNT > MLX5E_XDP_MPW_MAX_NUM_DS;
+}
+
+static inline void
+mlx5e_fill_xdpsq_frag_edge(struct mlx5e_xdpsq *sq, struct mlx5_wq_cyc *wq,
+			   u16 pi, u16 nnops)
+{
+	struct mlx5e_xdp_wqe_info *edge_wi, *wi = &sq->db.wqe_info[pi];
+
+	edge_wi = wi + nnops;
+	/* fill sq frag edge with nops to avoid wqe wrapping two pages */
+	for (; wi < edge_wi; wi++) {
+		wi->num_wqebbs = 1;
+		wi->num_pkts   = 0;
+		mlx5e_post_nop(wq, sq->sqn, &sq->pc);
+	}
+
+	sq->stats->nops += nnops;
+}
+
 static inline void
 mlx5e_xdp_mpwqe_add_dseg(struct mlx5e_xdpsq *sq,
 			 struct mlx5e_xdp_xmit_data *xdptxd,
@@ -126,20 +170,12 @@ mlx5e_xdp_mpwqe_add_dseg(struct mlx5e_xdpsq *sq,
 
 	session->pkt_count++;
 
-#define MLX5E_XDP_INLINE_WQE_SZ_THRSD (256 - sizeof(struct mlx5_wqe_inline_seg))
-
 	if (session->inline_on && dma_len <= MLX5E_XDP_INLINE_WQE_SZ_THRSD) {
 		struct mlx5_wqe_inline_seg *inline_dseg =
 			(struct mlx5_wqe_inline_seg *)dseg;
 		u16 ds_len = sizeof(*inline_dseg) + dma_len;
 		u16 ds_cnt = DIV_ROUND_UP(ds_len, MLX5_SEND_WQE_DS);
 
-		if (unlikely(session->ds_count + ds_cnt > session->max_ds_count)) {
-			/* Not enough space for inline wqe, send with memory pointer */
-			session->complete = true;
-			goto no_inline;
-		}
-
 		inline_dseg->byte_count = cpu_to_be32(dma_len | MLX5_INLINE_SEG);
 		memcpy(inline_dseg->data, xdptxd->data, dma_len);
 
@@ -148,7 +184,6 @@ mlx5e_xdp_mpwqe_add_dseg(struct mlx5e_xdpsq *sq,
 		return;
 	}
 
-no_inline:
 	dseg->addr       = cpu_to_be64(xdptxd->dma_addr);
 	dseg->byte_count = cpu_to_be32(dma_len);
 	dseg->lkey       = sq->mkey_be;
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_stats.c b/drivers/net/ethernet/mellanox/mlx5/core/en_stats.c
index 539b4d3656da..6eee3c7d4b06 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_stats.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_stats.c
@@ -74,6 +74,7 @@ static const struct counter_desc sw_stats_desc[] = {
 	{ MLX5E_DECLARE_STAT(struct mlx5e_sw_stats, rx_xdp_tx_xmit) },
 	{ MLX5E_DECLARE_STAT(struct mlx5e_sw_stats, rx_xdp_tx_mpwqe) },
 	{ MLX5E_DECLARE_STAT(struct mlx5e_sw_stats, rx_xdp_tx_inlnw) },
+	{ MLX5E_DECLARE_STAT(struct mlx5e_sw_stats, rx_xdp_tx_nops) },
 	{ MLX5E_DECLARE_STAT(struct mlx5e_sw_stats, rx_xdp_tx_full) },
 	{ MLX5E_DECLARE_STAT(struct mlx5e_sw_stats, rx_xdp_tx_err) },
 	{ MLX5E_DECLARE_STAT(struct mlx5e_sw_stats, rx_xdp_tx_cqe) },
@@ -90,6 +91,7 @@ static const struct counter_desc sw_stats_desc[] = {
 	{ MLX5E_DECLARE_STAT(struct mlx5e_sw_stats, tx_xdp_xmit) },
 	{ MLX5E_DECLARE_STAT(struct mlx5e_sw_stats, tx_xdp_mpwqe) },
 	{ MLX5E_DECLARE_STAT(struct mlx5e_sw_stats, tx_xdp_inlnw) },
+	{ MLX5E_DECLARE_STAT(struct mlx5e_sw_stats, tx_xdp_nops) },
 	{ MLX5E_DECLARE_STAT(struct mlx5e_sw_stats, tx_xdp_full) },
 	{ MLX5E_DECLARE_STAT(struct mlx5e_sw_stats, tx_xdp_err) },
 	{ MLX5E_DECLARE_STAT(struct mlx5e_sw_stats, tx_xdp_cqes) },
@@ -200,6 +202,7 @@ static void mlx5e_grp_sw_update_stats(struct mlx5e_priv *priv)
 		s->rx_xdp_tx_xmit  += xdpsq_stats->xmit;
 		s->rx_xdp_tx_mpwqe += xdpsq_stats->mpwqe;
 		s->rx_xdp_tx_inlnw += xdpsq_stats->inlnw;
+		s->rx_xdp_tx_nops  += xdpsq_stats->nops;
 		s->rx_xdp_tx_full  += xdpsq_stats->full;
 		s->rx_xdp_tx_err   += xdpsq_stats->err;
 		s->rx_xdp_tx_cqe   += xdpsq_stats->cqes;
@@ -227,6 +230,7 @@ static void mlx5e_grp_sw_update_stats(struct mlx5e_priv *priv)
 		s->tx_xdp_xmit    += xdpsq_red_stats->xmit;
 		s->tx_xdp_mpwqe   += xdpsq_red_stats->mpwqe;
 		s->tx_xdp_inlnw   += xdpsq_red_stats->inlnw;
+		s->tx_xdp_nops	  += xdpsq_red_stats->nops;
 		s->tx_xdp_full    += xdpsq_red_stats->full;
 		s->tx_xdp_err     += xdpsq_red_stats->err;
 		s->tx_xdp_cqes    += xdpsq_red_stats->cqes;
@@ -1331,6 +1335,7 @@ static const struct counter_desc rq_xdpsq_stats_desc[] = {
 	{ MLX5E_DECLARE_RQ_XDPSQ_STAT(struct mlx5e_xdpsq_stats, xmit) },
 	{ MLX5E_DECLARE_RQ_XDPSQ_STAT(struct mlx5e_xdpsq_stats, mpwqe) },
 	{ MLX5E_DECLARE_RQ_XDPSQ_STAT(struct mlx5e_xdpsq_stats, inlnw) },
+	{ MLX5E_DECLARE_RQ_XDPSQ_STAT(struct mlx5e_xdpsq_stats, nops) },
 	{ MLX5E_DECLARE_RQ_XDPSQ_STAT(struct mlx5e_xdpsq_stats, full) },
 	{ MLX5E_DECLARE_RQ_XDPSQ_STAT(struct mlx5e_xdpsq_stats, err) },
 	{ MLX5E_DECLARE_RQ_XDPSQ_STAT(struct mlx5e_xdpsq_stats, cqes) },
@@ -1340,6 +1345,7 @@ static const struct counter_desc xdpsq_stats_desc[] = {
 	{ MLX5E_DECLARE_XDPSQ_STAT(struct mlx5e_xdpsq_stats, xmit) },
 	{ MLX5E_DECLARE_XDPSQ_STAT(struct mlx5e_xdpsq_stats, mpwqe) },
 	{ MLX5E_DECLARE_XDPSQ_STAT(struct mlx5e_xdpsq_stats, inlnw) },
+	{ MLX5E_DECLARE_XDPSQ_STAT(struct mlx5e_xdpsq_stats, nops) },
 	{ MLX5E_DECLARE_XDPSQ_STAT(struct mlx5e_xdpsq_stats, full) },
 	{ MLX5E_DECLARE_XDPSQ_STAT(struct mlx5e_xdpsq_stats, err) },
 	{ MLX5E_DECLARE_XDPSQ_STAT(struct mlx5e_xdpsq_stats, cqes) },
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_stats.h b/drivers/net/ethernet/mellanox/mlx5/core/en_stats.h
index 76ac111e14d0..bf645d42c833 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_stats.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_stats.h
@@ -81,6 +81,7 @@ struct mlx5e_sw_stats {
 	u64 rx_xdp_tx_xmit;
 	u64 rx_xdp_tx_mpwqe;
 	u64 rx_xdp_tx_inlnw;
+	u64 rx_xdp_tx_nops;
 	u64 rx_xdp_tx_full;
 	u64 rx_xdp_tx_err;
 	u64 rx_xdp_tx_cqe;
@@ -97,6 +98,7 @@ struct mlx5e_sw_stats {
 	u64 tx_xdp_xmit;
 	u64 tx_xdp_mpwqe;
 	u64 tx_xdp_inlnw;
+	u64 tx_xdp_nops;
 	u64 tx_xdp_full;
 	u64 tx_xdp_err;
 	u64 tx_xdp_cqes;
@@ -288,6 +290,7 @@ struct mlx5e_xdpsq_stats {
 	u64 xmit;
 	u64 mpwqe;
 	u64 inlnw;
+	u64 nops;
 	u64 full;
 	u64 err;
 	/* dirtied @completion */
-- 
2.21.0

[net-next 06/12] net/mlx5e: XDP, Slight enhancement for WQE fetch function

[Saeed Mahameed]

From: Tariq Toukan <tariqt@mellanox.com>

Instead of passing an output param, let function return the
WQE pointer.
In addition, pass &pi so it gets its value in the function,
and save the redundant assignment that comes after it.

Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
---
 drivers/net/ethernet/mellanox/mlx5/core/en/xdp.c |  4 +---
 drivers/net/ethernet/mellanox/mlx5/core/en/xdp.h | 13 ++++++++-----
 2 files changed, 9 insertions(+), 8 deletions(-)

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.c b/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.c
index 8cb98326531f..1ed5c33e022f 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.c
@@ -187,14 +187,12 @@ static void mlx5e_xdp_mpwqe_session_start(struct mlx5e_xdpsq *sq)
 	if (unlikely(contig_wqebbs < MLX5_SEND_WQE_MAX_WQEBBS))
 		mlx5e_fill_xdpsq_frag_edge(sq, wq, pi, contig_wqebbs);
 
-	mlx5e_xdpsq_fetch_wqe(sq, &session->wqe);
+	session->wqe = mlx5e_xdpsq_fetch_wqe(sq, &pi);
 
 	prefetchw(session->wqe->data);
 	session->ds_count  = MLX5E_XDP_TX_EMPTY_DS_COUNT;
 	session->pkt_count = 0;
 
-	pi = mlx5_wq_cyc_ctr2ix(wq, sq->pc);
-
 	mlx5e_xdp_update_inline_state(sq);
 
 	stats->mpwqe++;
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.h b/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.h
index e0ed7710f5f1..36ac1e3816b9 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.h
@@ -190,14 +190,17 @@ mlx5e_xdp_mpwqe_add_dseg(struct mlx5e_xdpsq *sq,
 	session->ds_count++;
 }
 
-static inline void mlx5e_xdpsq_fetch_wqe(struct mlx5e_xdpsq *sq,
-					 struct mlx5e_tx_wqe **wqe)
+static inline struct mlx5e_tx_wqe *
+mlx5e_xdpsq_fetch_wqe(struct mlx5e_xdpsq *sq, u16 *pi)
 {
 	struct mlx5_wq_cyc *wq = &sq->wq;
-	u16 pi = mlx5_wq_cyc_ctr2ix(wq, sq->pc);
+	struct mlx5e_tx_wqe *wqe;
 
-	*wqe = mlx5_wq_cyc_get_wqe(wq, pi);
-	memset(*wqe, 0, sizeof(**wqe));
+	*pi = mlx5_wq_cyc_ctr2ix(wq, sq->pc);
+	wqe = mlx5_wq_cyc_get_wqe(wq, *pi);
+	memset(wqe, 0, sizeof(*wqe));
+
+	return wqe;
 }
 
 static inline void
-- 
2.21.0

[net-next 07/12] net/mlx5e: Tx, Soften inline mode VLAN dependencies

[Saeed Mahameed]

From: Tariq Toukan <tariqt@mellanox.com>

If capable, use zero inline mode in TX WQE for non-VLAN packets.
For VLAN ones, keep the enforcement of at least L2 inline mode,
unless the WQE VLAN insertion offload cap is on.

Performance:
Tested single core packet rate of 64Bytes.

NIC: ConnectX-5
CPU: Intel(R) Xeon(R) Gold 6154 CPU @ 3.00GHz

pktgen:
Before: 12.46 Mpps
After:  14.65 Mpps (+17.5%)

XDP_TX:
The MPWQE flow is not affected, as it already has this optimization.
So we test with priv-flag xdp_tx_mpwqe: off.

Before:  9.90 Mpps
After:  10.20 Mpps (+3%)

Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Tested-by: Noam Stolero <noams@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
---
 drivers/net/ethernet/mellanox/mlx5/core/en.h  |  2 +-
 .../net/ethernet/mellanox/mlx5/core/en/txrx.h | 22 +++++++++++++++++--
 .../ethernet/mellanox/mlx5/core/en_common.c   | 12 ----------
 .../ethernet/mellanox/mlx5/core/en_dcbnl.c    |  2 +-
 .../net/ethernet/mellanox/mlx5/core/en_main.c |  4 +++-
 .../net/ethernet/mellanox/mlx5/core/en_tx.c   |  7 +++---
 .../net/ethernet/mellanox/mlx5/core/vport.c   |  7 +++---
 7 files changed, 33 insertions(+), 23 deletions(-)

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en.h b/drivers/net/ethernet/mellanox/mlx5/core/en.h
index 745bcc25c6f8..30f13f81c965 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en.h
@@ -359,6 +359,7 @@ enum {
 	MLX5E_SQ_STATE_IPSEC,
 	MLX5E_SQ_STATE_AM,
 	MLX5E_SQ_STATE_TLS,
+	MLX5E_SQ_STATE_VLAN_NEED_L2_INLINE,
 };
 
 struct mlx5e_sq_wqe_info {
@@ -1132,7 +1133,6 @@ void mlx5e_build_rq_params(struct mlx5_core_dev *mdev,
 			   struct mlx5e_params *params);
 void mlx5e_build_rss_params(struct mlx5e_rss_params *rss_params,
 			    u16 num_channels);
-u8 mlx5e_params_calculate_tx_min_inline(struct mlx5_core_dev *mdev);
 void mlx5e_rx_dim_work(struct work_struct *work);
 void mlx5e_tx_dim_work(struct work_struct *work);
 
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/txrx.h b/drivers/net/ethernet/mellanox/mlx5/core/en/txrx.h
index 7da22b413a48..87be96747902 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en/txrx.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en/txrx.h
@@ -117,9 +117,27 @@ mlx5e_notify_hw(struct mlx5_wq_cyc *wq, u16 pc, void __iomem *uar_map,
 	mlx5_write64((__be32 *)ctrl, uar_map);
 }
 
-static inline bool mlx5e_transport_inline_tx_wqe(struct mlx5e_tx_wqe *wqe)
+static inline bool mlx5e_transport_inline_tx_wqe(struct mlx5_wqe_ctrl_seg *cseg)
 {
-	return !!wqe->ctrl.tisn;
+	return cseg && !!cseg->tisn;
+}
+
+static inline u8
+mlx5e_tx_wqe_inline_mode(struct mlx5e_txqsq *sq, struct mlx5_wqe_ctrl_seg *cseg,
+			 struct sk_buff *skb)
+{
+	u8 mode;
+
+	if (mlx5e_transport_inline_tx_wqe(cseg))
+		return MLX5_INLINE_MODE_TCP_UDP;
+
+	mode = sq->min_inline_mode;
+
+	if (skb_vlan_tag_present(skb) &&
+	    test_bit(MLX5E_SQ_STATE_VLAN_NEED_L2_INLINE, &sq->state))
+		mode = max_t(u8, MLX5_INLINE_MODE_L2, mode);
+
+	return mode;
 }
 
 static inline void mlx5e_cq_arm(struct mlx5e_cq *cq)
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_common.c b/drivers/net/ethernet/mellanox/mlx5/core/en_common.c
index 1539cf3de5dc..f7890e0ce96c 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_common.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_common.c
@@ -180,15 +180,3 @@ int mlx5e_refresh_tirs(struct mlx5e_priv *priv, bool enable_uc_lb)
 
 	return err;
 }
-
-u8 mlx5e_params_calculate_tx_min_inline(struct mlx5_core_dev *mdev)
-{
-	u8 min_inline_mode;
-
-	mlx5_query_min_inline(mdev, &min_inline_mode);
-	if (min_inline_mode == MLX5_INLINE_MODE_NONE &&
-	    !MLX5_CAP_ETH(mdev, wqe_vlan_insert))
-		min_inline_mode = MLX5_INLINE_MODE_L2;
-
-	return min_inline_mode;
-}
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_dcbnl.c b/drivers/net/ethernet/mellanox/mlx5/core/en_dcbnl.c
index 8dd31b5c740c..01f2918063af 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_dcbnl.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_dcbnl.c
@@ -1101,7 +1101,7 @@ void mlx5e_dcbnl_delete_app(struct mlx5e_priv *priv)
 static void mlx5e_trust_update_tx_min_inline_mode(struct mlx5e_priv *priv,
 						  struct mlx5e_params *params)
 {
-	params->tx_min_inline_mode = mlx5e_params_calculate_tx_min_inline(priv->mdev);
+	mlx5_query_min_inline(priv->mdev, &params->tx_min_inline_mode);
 	if (priv->dcbx_dp.trust_state == MLX5_QPTS_TRUST_DSCP &&
 	    params->tx_min_inline_mode == MLX5_INLINE_MODE_L2)
 		params->tx_min_inline_mode = MLX5_INLINE_MODE_IP;
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_main.c b/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
index b2618dd6dd10..e75cb18c2256 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
@@ -1131,6 +1131,8 @@ static int mlx5e_alloc_txqsq(struct mlx5e_channel *c,
 	sq->stats     = &c->priv->channel_stats[c->ix].sq[tc];
 	sq->stop_room = MLX5E_SQ_STOP_ROOM;
 	INIT_WORK(&sq->recover_work, mlx5e_tx_err_cqe_work);
+	if (!MLX5_CAP_ETH(mdev, wqe_vlan_insert))
+		set_bit(MLX5E_SQ_STATE_VLAN_NEED_L2_INLINE, &sq->state);
 	if (MLX5_IPSEC_DEV(c->priv->mdev))
 		set_bit(MLX5E_SQ_STATE_IPSEC, &sq->state);
 	if (mlx5_accel_is_tls_device(c->priv->mdev)) {
@@ -4777,7 +4779,7 @@ void mlx5e_build_nic_params(struct mlx5_core_dev *mdev,
 	mlx5e_set_tx_cq_mode_params(params, MLX5_CQ_PERIOD_MODE_START_FROM_EQE);
 
 	/* TX inline */
-	params->tx_min_inline_mode = mlx5e_params_calculate_tx_min_inline(mdev);
+	mlx5_query_min_inline(mdev, &params->tx_min_inline_mode);
 
 	/* RSS */
 	mlx5e_build_rss_params(rss_params, params->num_channels);
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_tx.c b/drivers/net/ethernet/mellanox/mlx5/core/en_tx.c
index acf25cc38fa1..d3a67a9b4eba 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_tx.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_tx.c
@@ -292,8 +292,7 @@ netdev_tx_t mlx5e_sq_xmit(struct mlx5e_txqsq *sq, struct sk_buff *skb,
 		num_bytes = skb->len + (skb_shinfo(skb)->gso_segs - 1) * ihs;
 		stats->packets += skb_shinfo(skb)->gso_segs;
 	} else {
-		u8 mode = mlx5e_transport_inline_tx_wqe(wqe) ?
-			MLX5_INLINE_MODE_TCP_UDP : sq->min_inline_mode;
+		u8 mode = mlx5e_tx_wqe_inline_mode(sq, &wqe->ctrl, skb);
 
 		opcode    = MLX5_OPCODE_SEND;
 		mss       = 0;
@@ -608,9 +607,11 @@ netdev_tx_t mlx5i_sq_xmit(struct mlx5e_txqsq *sq, struct sk_buff *skb,
 		num_bytes = skb->len + (skb_shinfo(skb)->gso_segs - 1) * ihs;
 		stats->packets += skb_shinfo(skb)->gso_segs;
 	} else {
+		u8 mode = mlx5e_tx_wqe_inline_mode(sq, NULL, skb);
+
 		opcode    = MLX5_OPCODE_SEND;
 		mss       = 0;
-		ihs       = mlx5e_calc_min_inline(sq->min_inline_mode, skb);
+		ihs       = mlx5e_calc_min_inline(mode, skb);
 		num_bytes = max_t(unsigned int, skb->len, ETH_ZLEN);
 		stats->packets++;
 	}
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/vport.c b/drivers/net/ethernet/mellanox/mlx5/core/vport.c
index c912d82ca64b..30f7848a6f88 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/vport.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/vport.c
@@ -122,12 +122,13 @@ void mlx5_query_min_inline(struct mlx5_core_dev *mdev,
 			   u8 *min_inline_mode)
 {
 	switch (MLX5_CAP_ETH(mdev, wqe_inline_mode)) {
+	case MLX5_CAP_INLINE_MODE_VPORT_CONTEXT:
+		if (!mlx5_query_nic_vport_min_inline(mdev, 0, min_inline_mode))
+			break;
+		/* fall through */
 	case MLX5_CAP_INLINE_MODE_L2:
 		*min_inline_mode = MLX5_INLINE_MODE_L2;
 		break;
-	case MLX5_CAP_INLINE_MODE_VPORT_CONTEXT:
-		mlx5_query_nic_vport_min_inline(mdev, 0, min_inline_mode);
-		break;
 	case MLX5_CAP_INLINE_MODE_NOT_REQUIRED:
 		*min_inline_mode = MLX5_INLINE_MODE_NONE;
 		break;
-- 
2.21.0

[net-next 08/12] net/mlx5e: Rx, checksum handling refactoring

[Saeed Mahameed]

Move vlan checksum fixup flow into mlx5e_skb_padding_csum(), which is
supposed to fixup SKB checksum if needed. And rename
mlx5e_skb_padding_csum() to mlx5e_skb_csum_fixup().

Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
---
 .../net/ethernet/mellanox/mlx5/core/en_rx.c   | 26 ++++++++++---------
 1 file changed, 14 insertions(+), 12 deletions(-)

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_rx.c b/drivers/net/ethernet/mellanox/mlx5/core/en_rx.c
index ac6e586d403d..60570b442fff 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_rx.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_rx.c
@@ -859,13 +859,24 @@ tail_padding_csum(struct sk_buff *skb, int offset,
 }
 
 static void
-mlx5e_skb_padding_csum(struct sk_buff *skb, int network_depth, __be16 proto,
-		       struct mlx5e_rq_stats *stats)
+mlx5e_skb_csum_fixup(struct sk_buff *skb, int network_depth, __be16 proto,
+		     struct mlx5e_rq_stats *stats)
 {
 	struct ipv6hdr *ip6;
 	struct iphdr   *ip4;
 	int pkt_len;
 
+	/* Fixup vlan headers, if any */
+	if (network_depth > ETH_HLEN)
+		/* CQE csum is calculated from the IP header and does
+		 * not cover VLAN headers (if present). This will add
+		 * the checksum manually.
+		 */
+		skb->csum = csum_partial(skb->data + ETH_HLEN,
+					 network_depth - ETH_HLEN,
+					 skb->csum);
+
+	/* Fixup tail padding, if any */
 	switch (proto) {
 	case htons(ETH_P_IP):
 		ip4 = (struct iphdr *)(skb->data + network_depth);
@@ -931,16 +942,7 @@ static inline void mlx5e_handle_csum(struct net_device *netdev,
 			return; /* CQE csum covers all received bytes */
 
 		/* csum might need some fixups ...*/
-		if (network_depth > ETH_HLEN)
-			/* CQE csum is calculated from the IP header and does
-			 * not cover VLAN headers (if present). This will add
-			 * the checksum manually.
-			 */
-			skb->csum = csum_partial(skb->data + ETH_HLEN,
-						 network_depth - ETH_HLEN,
-						 skb->csum);
-
-		mlx5e_skb_padding_csum(skb, network_depth, proto, stats);
+		mlx5e_skb_csum_fixup(skb, network_depth, proto, stats);
 		return;
 	}
 
-- 
2.21.0

[net-next 09/12] net/mlx5e: Fix mlx5e_tx_reporter_create return value

[Saeed Mahameed]

From: Aya Levin <ayal@mellanox.com>

Return error when failing to create a reporter in devlink. Since
NET_DEVLINK mandatory to MLX5_CORE in Kconfig, returned pointer
can't be NULL and can only hold an error in bad path.

Signed-off-by: Aya Levin <ayal@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
---
 drivers/net/ethernet/mellanox/mlx5/core/en/reporter_tx.c | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/reporter_tx.c b/drivers/net/ethernet/mellanox/mlx5/core/en/reporter_tx.c
index f3d98748b211..383ecfd85d8a 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en/reporter_tx.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en/reporter_tx.c
@@ -296,11 +296,13 @@ int mlx5e_tx_reporter_create(struct mlx5e_priv *priv)
 		devlink_health_reporter_create(devlink, &mlx5_tx_reporter_ops,
 					       MLX5_REPORTER_TX_GRACEFUL_PERIOD,
 					       true, priv);
-	if (IS_ERR(priv->tx_reporter))
+	if (IS_ERR(priv->tx_reporter)) {
 		netdev_warn(priv->netdev,
 			    "Failed to create tx reporter, err = %ld\n",
 			    PTR_ERR(priv->tx_reporter));
-	return IS_ERR_OR_NULL(priv->tx_reporter);
+		return PTR_ERR(priv->tx_reporter);
+	}
+	return 0;
 }
 
 void mlx5e_tx_reporter_destroy(struct mlx5e_priv *priv)
-- 
2.21.0

[net-next 10/12] net/mlx5e: Set tx reporter only on successful creation

[Saeed Mahameed]

From: Aya Levin <ayal@mellanox.com>

When failing to create tx reporter, don't set the reporter's pointer.
Creating a reporter is not mandatory for driver load, avoid
garbage/error pointer.

Signed-off-by: Aya Levin <ayal@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
---
 .../ethernet/mellanox/mlx5/core/en/reporter_tx.c   | 14 ++++++++------
 drivers/net/ethernet/mellanox/mlx5/core/en_main.c  |  2 +-
 2 files changed, 9 insertions(+), 7 deletions(-)

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/reporter_tx.c b/drivers/net/ethernet/mellanox/mlx5/core/en/reporter_tx.c
index 383ecfd85d8a..f1c652f75718 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en/reporter_tx.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en/reporter_tx.c
@@ -117,7 +117,7 @@ static int mlx5_tx_health_report(struct devlink_health_reporter *tx_reporter,
 				 char *err_str,
 				 struct mlx5e_tx_err_ctx *err_ctx)
 {
-	if (IS_ERR_OR_NULL(tx_reporter)) {
+	if (!tx_reporter) {
 		netdev_err(err_ctx->sq->channel->netdev, err_str);
 		return err_ctx->recover(err_ctx->sq);
 	}
@@ -289,25 +289,27 @@ static const struct devlink_health_reporter_ops mlx5_tx_reporter_ops = {
 
 int mlx5e_tx_reporter_create(struct mlx5e_priv *priv)
 {
+	struct devlink_health_reporter *reporter;
 	struct mlx5_core_dev *mdev = priv->mdev;
 	struct devlink *devlink = priv_to_devlink(mdev);
 
-	priv->tx_reporter =
+	reporter =
 		devlink_health_reporter_create(devlink, &mlx5_tx_reporter_ops,
 					       MLX5_REPORTER_TX_GRACEFUL_PERIOD,
 					       true, priv);
-	if (IS_ERR(priv->tx_reporter)) {
+	if (IS_ERR(reporter)) {
 		netdev_warn(priv->netdev,
 			    "Failed to create tx reporter, err = %ld\n",
-			    PTR_ERR(priv->tx_reporter));
-		return PTR_ERR(priv->tx_reporter);
+			    PTR_ERR(reporter));
+		return PTR_ERR(reporter);
 	}
+	priv->tx_reporter = reporter;
 	return 0;
 }
 
 void mlx5e_tx_reporter_destroy(struct mlx5e_priv *priv)
 {
-	if (IS_ERR_OR_NULL(priv->tx_reporter))
+	if (!priv->tx_reporter)
 		return;
 
 	devlink_health_reporter_destroy(priv->tx_reporter);
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_main.c b/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
index e75cb18c2256..4db595a7eb03 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
@@ -2325,7 +2325,7 @@ int mlx5e_open_channels(struct mlx5e_priv *priv,
 			goto err_close_channels;
 	}
 
-	if (!IS_ERR_OR_NULL(priv->tx_reporter))
+	if (priv->tx_reporter)
 		devlink_health_reporter_state_update(priv->tx_reporter,
 						     DEVLINK_HEALTH_REPORTER_STATE_HEALTHY);
 
-- 
2.21.0

[net-next 11/12] net/mlx5e: TX reporter cleanup

[Saeed Mahameed]

From: Aya Levin <ayal@mellanox.com>

Remove redundant include files.

Signed-off-by: Aya Levin <ayal@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
---
 drivers/net/ethernet/mellanox/mlx5/core/en/reporter.h    | 1 -
 drivers/net/ethernet/mellanox/mlx5/core/en/reporter_tx.c | 1 -
 2 files changed, 2 deletions(-)

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/reporter.h b/drivers/net/ethernet/mellanox/mlx5/core/en/reporter.h
index e78e92753d73..ed7a3881d2c5 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en/reporter.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en/reporter.h
@@ -4,7 +4,6 @@
 #ifndef __MLX5E_EN_REPORTER_H
 #define __MLX5E_EN_REPORTER_H
 
-#include <linux/mlx5/driver.h>
 #include "en.h"
 
 int mlx5e_tx_reporter_create(struct mlx5e_priv *priv);
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/reporter_tx.c b/drivers/net/ethernet/mellanox/mlx5/core/en/reporter_tx.c
index f1c652f75718..6e54fefea410 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en/reporter_tx.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en/reporter_tx.c
@@ -1,7 +1,6 @@
 /* SPDX-License-Identifier: GPL-2.0 */
 /* Copyright (c) 2019 Mellanox Technologies. */
 
-#include <net/devlink.h>
 #include "reporter.h"
 #include "lib/eq.h"
 
-- 
2.21.0

[net-next 12/12] net/mlx5e: Allow dropping specific tunnel packets

[Saeed Mahameed]

From: Tonghao Zhang <xiangxia.m.yue@gmail.com>

In some case, we don't want to allow specific tunnel packets
to host that can avoid to take up high CPU (e.g network attacks).
But other tunnel packets which not matched in hardware will be
sent to host too.

    $ tc filter add dev vxlan_sys_4789 \
	    protocol ip chain 0 parent ffff: prio 1 handle 1 \
	    flower dst_ip 1.1.1.100 ip_proto tcp dst_port 80 \
	    enc_dst_ip 2.2.2.100 enc_key_id 100 enc_dst_port 4789 \
	    action tunnel_key unset pipe action drop

Signed-off-by: Tonghao Zhang <xiangxia.m.yue@gmail.com>
Reviewed-by: Roi Dayan <roid@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
---
 drivers/net/ethernet/mellanox/mlx5/core/en_tc.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_tc.c b/drivers/net/ethernet/mellanox/mlx5/core/en_tc.c
index dc5fc3350b65..c5d75e2ecf54 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_tc.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_tc.c
@@ -2485,7 +2485,8 @@ static bool actions_match_supported(struct mlx5e_priv *priv,
 
 	if (flow_flag_test(flow, EGRESS) &&
 	    !((actions & MLX5_FLOW_CONTEXT_ACTION_DECAP) ||
-	      (actions & MLX5_FLOW_CONTEXT_ACTION_VLAN_POP)))
+	      (actions & MLX5_FLOW_CONTEXT_ACTION_VLAN_POP) ||
+	      (actions & MLX5_FLOW_CONTEXT_ACTION_DROP)))
 		return false;
 
 	if (actions & MLX5_FLOW_CONTEXT_ACTION_MOD_HDR)
-- 
2.21.0

Re: [pull request][net-next 00/12] Mellanox, mlx5 updates 2019-08-01

[David Miller]

From: Saeed Mahameed <saeedm@mellanox.com>
Date: Thu, 1 Aug 2019 19:56:50 +0000

> This series provides misc updates to mlx5 driver.
> For more information please see tag log below.
> 
> Please pull and let me know if there is any problem.
> 
> Please note that the series starts with a merge of mlx5-next branch,
> to resolve and avoid dependency with rdma tree.

Pulled, thank you.