* [net 1/5] net/mlx5: Lag, change multipath and bonding to be mutually exclusive
2021-10-20 17:56 [pull request][net 0/5] mlx5 fixes 2021-10-20 Saeed Mahameed
@ 2021-10-20 17:56 ` Saeed Mahameed
2021-10-21 11:20 ` patchwork-bot+netdevbpf
2021-10-20 17:56 ` [net 2/5] net/mlx5: E-switch, Return correct error code on group creation failure Saeed Mahameed
` (3 subsequent siblings)
4 siblings, 1 reply; 7+ messages in thread
From: Saeed Mahameed @ 2021-10-20 17:56 UTC (permalink / raw)
To: David S. Miller, Jakub Kicinski
Cc: netdev, Maor Dickman, Roi Dayan, Mark Bloch, Saeed Mahameed
From: Maor Dickman <maord@nvidia.com>
Both multipath and bonding events are changing the HW LAG state
independently.
Handling one of the features events while the other is already
enabled can cause unwanted behavior, for example handling
bonding event while multipath enabled will disable the lag and
cause multipath to stop working.
Fix it by ignoring bonding event while in multipath and ignoring FIB
events while in bonding mode.
Fixes: 544fe7c2e654 ("net/mlx5e: Activate HW multipath and handle port affinity based on FIB events")
Signed-off-by: Maor Dickman <maord@nvidia.com>
Reviewed-by: Roi Dayan <roid@nvidia.com>
Reviewed-by: Mark Bloch <mbloch@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
---
drivers/net/ethernet/mellanox/mlx5/core/en/tc_tun.c | 2 ++
drivers/net/ethernet/mellanox/mlx5/core/en_tc.c | 2 ++
drivers/net/ethernet/mellanox/mlx5/core/lag.c | 4 ++++
drivers/net/ethernet/mellanox/mlx5/core/lag_mp.c | 13 ++++++++-----
drivers/net/ethernet/mellanox/mlx5/core/lag_mp.h | 2 ++
include/linux/mlx5/driver.h | 1 -
6 files changed, 18 insertions(+), 6 deletions(-)
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/tc_tun.c b/drivers/net/ethernet/mellanox/mlx5/core/en/tc_tun.c
index b4e986818794..4a13ef561587 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en/tc_tun.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en/tc_tun.c
@@ -10,6 +10,8 @@
#include "en_tc.h"
#include "rep/tc.h"
#include "rep/neigh.h"
+#include "lag.h"
+#include "lag_mp.h"
struct mlx5e_tc_tun_route_attr {
struct net_device *out_dev;
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_tc.c b/drivers/net/ethernet/mellanox/mlx5/core/en_tc.c
index ba8164792016..129ff7e0d65c 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_tc.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_tc.c
@@ -67,6 +67,8 @@
#include "lib/fs_chains.h"
#include "diag/en_tc_tracepoint.h"
#include <asm/div64.h>
+#include "lag.h"
+#include "lag_mp.h"
#define nic_chains(priv) ((priv)->fs.tc.chains)
#define MLX5_MH_ACT_SZ MLX5_UN_SZ_BYTES(set_add_copy_action_in_auto)
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/lag.c b/drivers/net/ethernet/mellanox/mlx5/core/lag.c
index ca5690b0a7ab..d2105c1635c3 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/lag.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/lag.c
@@ -442,6 +442,10 @@ static void mlx5_do_bond(struct mlx5_lag *ldev)
if (!mlx5_lag_is_ready(ldev)) {
do_bond = false;
} else {
+ /* VF LAG is in multipath mode, ignore bond change requests */
+ if (mlx5_lag_is_multipath(dev0))
+ return;
+
tracker = ldev->tracker;
do_bond = tracker.is_bonded && mlx5_lag_check_prereq(ldev);
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/lag_mp.c b/drivers/net/ethernet/mellanox/mlx5/core/lag_mp.c
index f239b352a58a..21fdaf708f1f 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/lag_mp.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/lag_mp.c
@@ -9,20 +9,23 @@
#include "eswitch.h"
#include "lib/mlx5.h"
+static bool __mlx5_lag_is_multipath(struct mlx5_lag *ldev)
+{
+ return !!(ldev->flags & MLX5_LAG_FLAG_MULTIPATH);
+}
+
static bool mlx5_lag_multipath_check_prereq(struct mlx5_lag *ldev)
{
if (!mlx5_lag_is_ready(ldev))
return false;
+ if (__mlx5_lag_is_active(ldev) && !__mlx5_lag_is_multipath(ldev))
+ return false;
+
return mlx5_esw_multipath_prereq(ldev->pf[MLX5_LAG_P1].dev,
ldev->pf[MLX5_LAG_P2].dev);
}
-static bool __mlx5_lag_is_multipath(struct mlx5_lag *ldev)
-{
- return !!(ldev->flags & MLX5_LAG_FLAG_MULTIPATH);
-}
-
bool mlx5_lag_is_multipath(struct mlx5_core_dev *dev)
{
struct mlx5_lag *ldev;
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/lag_mp.h b/drivers/net/ethernet/mellanox/mlx5/core/lag_mp.h
index 729c839397a8..dea199e79bed 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/lag_mp.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/lag_mp.h
@@ -24,12 +24,14 @@ struct lag_mp {
void mlx5_lag_mp_reset(struct mlx5_lag *ldev);
int mlx5_lag_mp_init(struct mlx5_lag *ldev);
void mlx5_lag_mp_cleanup(struct mlx5_lag *ldev);
+bool mlx5_lag_is_multipath(struct mlx5_core_dev *dev);
#else /* CONFIG_MLX5_ESWITCH */
static inline void mlx5_lag_mp_reset(struct mlx5_lag *ldev) {};
static inline int mlx5_lag_mp_init(struct mlx5_lag *ldev) { return 0; }
static inline void mlx5_lag_mp_cleanup(struct mlx5_lag *ldev) {}
+bool mlx5_lag_is_multipath(struct mlx5_core_dev *dev) { return false; }
#endif /* CONFIG_MLX5_ESWITCH */
#endif /* __MLX5_LAG_MP_H__ */
diff --git a/include/linux/mlx5/driver.h b/include/linux/mlx5/driver.h
index e23417424373..f17d2101af7a 100644
--- a/include/linux/mlx5/driver.h
+++ b/include/linux/mlx5/driver.h
@@ -1138,7 +1138,6 @@ int mlx5_cmd_create_vport_lag(struct mlx5_core_dev *dev);
int mlx5_cmd_destroy_vport_lag(struct mlx5_core_dev *dev);
bool mlx5_lag_is_roce(struct mlx5_core_dev *dev);
bool mlx5_lag_is_sriov(struct mlx5_core_dev *dev);
-bool mlx5_lag_is_multipath(struct mlx5_core_dev *dev);
bool mlx5_lag_is_active(struct mlx5_core_dev *dev);
bool mlx5_lag_is_master(struct mlx5_core_dev *dev);
bool mlx5_lag_is_shared_fdb(struct mlx5_core_dev *dev);
--
2.31.1
^ permalink raw reply related [flat|nested] 7+ messages in thread* Re: [net 1/5] net/mlx5: Lag, change multipath and bonding to be mutually exclusive
2021-10-20 17:56 ` [net 1/5] net/mlx5: Lag, change multipath and bonding to be mutually exclusive Saeed Mahameed
@ 2021-10-21 11:20 ` patchwork-bot+netdevbpf
0 siblings, 0 replies; 7+ messages in thread
From: patchwork-bot+netdevbpf @ 2021-10-21 11:20 UTC (permalink / raw)
To: Saeed Mahameed; +Cc: davem, kuba, netdev, maord, roid, mbloch, saeedm
Hello:
This series was applied to netdev/net.git (master)
by Saeed Mahameed <saeedm@nvidia.com>:
On Wed, 20 Oct 2021 10:56:23 -0700 you wrote:
> From: Maor Dickman <maord@nvidia.com>
>
> Both multipath and bonding events are changing the HW LAG state
> independently.
> Handling one of the features events while the other is already
> enabled can cause unwanted behavior, for example handling
> bonding event while multipath enabled will disable the lag and
> cause multipath to stop working.
>
> [...]
Here is the summary with links:
- [net,1/5] net/mlx5: Lag, change multipath and bonding to be mutually exclusive
https://git.kernel.org/netdev/net/c/14fe2471c628
- [net,2/5] net/mlx5: E-switch, Return correct error code on group creation failure
https://git.kernel.org/netdev/net/c/a6f74333548f
- [net,3/5] net/mlx5e: Fix vlan data lost during suspend flow
https://git.kernel.org/netdev/net/c/68e66e1a69cd
- [net,4/5] net/mlx5e: IPsec: Fix a misuse of the software parser's fields
https://git.kernel.org/netdev/net/c/d10457f85d4a
- [net,5/5] net/mlx5e: IPsec: Fix work queue entry ethernet segment checksum flags
https://git.kernel.org/netdev/net/c/1d0003239401
You are awesome, thank you!
--
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html
^ permalink raw reply [flat|nested] 7+ messages in thread
* [net 2/5] net/mlx5: E-switch, Return correct error code on group creation failure
2021-10-20 17:56 [pull request][net 0/5] mlx5 fixes 2021-10-20 Saeed Mahameed
2021-10-20 17:56 ` [net 1/5] net/mlx5: Lag, change multipath and bonding to be mutually exclusive Saeed Mahameed
@ 2021-10-20 17:56 ` Saeed Mahameed
2021-10-20 17:56 ` [net 3/5] net/mlx5e: Fix vlan data lost during suspend flow Saeed Mahameed
` (2 subsequent siblings)
4 siblings, 0 replies; 7+ messages in thread
From: Saeed Mahameed @ 2021-10-20 17:56 UTC (permalink / raw)
To: David S. Miller, Jakub Kicinski
Cc: netdev, Dmytro Linkin, Dan Carpenter, Saeed Mahameed
From: Dmytro Linkin <dlinkin@nvidia.com>
Dan Carpenter report:
The patch f47e04eb96e0: "net/mlx5: E-switch, Allow setting share/max
tx rate limits of rate groups" from May 31, 2021, leads to the
following Smatch static checker warning:
drivers/net/ethernet/mellanox/mlx5/core/esw/qos.c:483 esw_qos_create_rate_group()
warn: passing zero to 'ERR_PTR'
If min rate normalization failed then error code may be overwritten to 0
if scheduling element destruction succeed. Ignore this value and always
return initial one.
Fixes: f47e04eb96e0 ("net/mlx5: E-switch, Allow setting share/max tx rate limits of rate groups")
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Dmytro Linkin <dlinkin@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
---
drivers/net/ethernet/mellanox/mlx5/core/esw/qos.c | 7 +++----
1 file changed, 3 insertions(+), 4 deletions(-)
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/esw/qos.c b/drivers/net/ethernet/mellanox/mlx5/core/esw/qos.c
index 985e305179d1..c6cc67cb4f6a 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/esw/qos.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/esw/qos.c
@@ -473,10 +473,9 @@ esw_qos_create_rate_group(struct mlx5_eswitch *esw, struct netlink_ext_ack *exta
err_min_rate:
list_del(&group->list);
- err = mlx5_destroy_scheduling_element_cmd(esw->dev,
- SCHEDULING_HIERARCHY_E_SWITCH,
- group->tsar_ix);
- if (err)
+ if (mlx5_destroy_scheduling_element_cmd(esw->dev,
+ SCHEDULING_HIERARCHY_E_SWITCH,
+ group->tsar_ix))
NL_SET_ERR_MSG_MOD(extack, "E-Switch destroy TSAR for group failed");
err_sched_elem:
kfree(group);
--
2.31.1
^ permalink raw reply related [flat|nested] 7+ messages in thread* [net 3/5] net/mlx5e: Fix vlan data lost during suspend flow
2021-10-20 17:56 [pull request][net 0/5] mlx5 fixes 2021-10-20 Saeed Mahameed
2021-10-20 17:56 ` [net 1/5] net/mlx5: Lag, change multipath and bonding to be mutually exclusive Saeed Mahameed
2021-10-20 17:56 ` [net 2/5] net/mlx5: E-switch, Return correct error code on group creation failure Saeed Mahameed
@ 2021-10-20 17:56 ` Saeed Mahameed
2021-10-20 17:56 ` [net 4/5] net/mlx5e: IPsec: Fix a misuse of the software parser's fields Saeed Mahameed
2021-10-20 17:56 ` [net 5/5] net/mlx5e: IPsec: Fix work queue entry ethernet segment checksum flags Saeed Mahameed
4 siblings, 0 replies; 7+ messages in thread
From: Saeed Mahameed @ 2021-10-20 17:56 UTC (permalink / raw)
To: David S. Miller, Jakub Kicinski
Cc: netdev, Moshe Shemesh, Tariq Toukan, Saeed Mahameed
From: Moshe Shemesh <moshe@nvidia.com>
During suspend flow the driver calls mlx5e_destroy_vlan_table() which
does not only delete the vlans steering flow rules, but also frees the
data on currently active vlans, thus it is not restored during resume
flow.
This fix keeps the vlan data on suspend flow and frees it only on driver
remove flow.
Fixes: 6783f0a21a3c ("net/mlx5e: Dynamic alloc vlan table for netdev when needed")
Signed-off-by: Moshe Shemesh <moshe@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
---
.../net/ethernet/mellanox/mlx5/core/en/fs.h | 3 ++
.../net/ethernet/mellanox/mlx5/core/en_fs.c | 28 +++++++++++--------
.../net/ethernet/mellanox/mlx5/core/en_main.c | 7 +++++
3 files changed, 26 insertions(+), 12 deletions(-)
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/fs.h b/drivers/net/ethernet/mellanox/mlx5/core/en/fs.h
index 41684a6c44e9..a88a1a48229f 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en/fs.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en/fs.h
@@ -199,6 +199,9 @@ void mlx5e_disable_cvlan_filter(struct mlx5e_priv *priv);
int mlx5e_create_flow_steering(struct mlx5e_priv *priv);
void mlx5e_destroy_flow_steering(struct mlx5e_priv *priv);
+int mlx5e_fs_init(struct mlx5e_priv *priv);
+void mlx5e_fs_cleanup(struct mlx5e_priv *priv);
+
int mlx5e_add_vlan_trap(struct mlx5e_priv *priv, int trap_id, int tir_num);
void mlx5e_remove_vlan_trap(struct mlx5e_priv *priv);
int mlx5e_add_mac_trap(struct mlx5e_priv *priv, int trap_id, int tir_num);
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_fs.c b/drivers/net/ethernet/mellanox/mlx5/core/en_fs.c
index c06b4b938ae7..d226cc5ab1d1 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_fs.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_fs.c
@@ -1186,10 +1186,6 @@ static int mlx5e_create_vlan_table(struct mlx5e_priv *priv)
struct mlx5e_flow_table *ft;
int err;
- priv->fs.vlan = kvzalloc(sizeof(*priv->fs.vlan), GFP_KERNEL);
- if (!priv->fs.vlan)
- return -ENOMEM;
-
ft = &priv->fs.vlan->ft;
ft->num_groups = 0;
@@ -1198,10 +1194,8 @@ static int mlx5e_create_vlan_table(struct mlx5e_priv *priv)
ft_attr.prio = MLX5E_NIC_PRIO;
ft->t = mlx5_create_flow_table(priv->fs.ns, &ft_attr);
- if (IS_ERR(ft->t)) {
- err = PTR_ERR(ft->t);
- goto err_free_t;
- }
+ if (IS_ERR(ft->t))
+ return PTR_ERR(ft->t);
ft->g = kcalloc(MLX5E_NUM_VLAN_GROUPS, sizeof(*ft->g), GFP_KERNEL);
if (!ft->g) {
@@ -1221,9 +1215,6 @@ static int mlx5e_create_vlan_table(struct mlx5e_priv *priv)
kfree(ft->g);
err_destroy_vlan_table:
mlx5_destroy_flow_table(ft->t);
-err_free_t:
- kvfree(priv->fs.vlan);
- priv->fs.vlan = NULL;
return err;
}
@@ -1232,7 +1223,6 @@ static void mlx5e_destroy_vlan_table(struct mlx5e_priv *priv)
{
mlx5e_del_vlan_rules(priv);
mlx5e_destroy_flow_table(&priv->fs.vlan->ft);
- kvfree(priv->fs.vlan);
}
static void mlx5e_destroy_inner_ttc_table(struct mlx5e_priv *priv)
@@ -1351,3 +1341,17 @@ void mlx5e_destroy_flow_steering(struct mlx5e_priv *priv)
mlx5e_arfs_destroy_tables(priv);
mlx5e_ethtool_cleanup_steering(priv);
}
+
+int mlx5e_fs_init(struct mlx5e_priv *priv)
+{
+ priv->fs.vlan = kvzalloc(sizeof(*priv->fs.vlan), GFP_KERNEL);
+ if (!priv->fs.vlan)
+ return -ENOMEM;
+ return 0;
+}
+
+void mlx5e_fs_cleanup(struct mlx5e_priv *priv)
+{
+ kvfree(priv->fs.vlan);
+ priv->fs.vlan = NULL;
+}
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_main.c b/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
index 09c8b71b186c..41ef6eb70a58 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
@@ -4578,6 +4578,12 @@ static int mlx5e_nic_init(struct mlx5_core_dev *mdev,
mlx5e_timestamp_init(priv);
+ err = mlx5e_fs_init(priv);
+ if (err) {
+ mlx5_core_err(mdev, "FS initialization failed, %d\n", err);
+ return err;
+ }
+
err = mlx5e_ipsec_init(priv);
if (err)
mlx5_core_err(mdev, "IPSec initialization failed, %d\n", err);
@@ -4595,6 +4601,7 @@ static void mlx5e_nic_cleanup(struct mlx5e_priv *priv)
mlx5e_health_destroy_reporters(priv);
mlx5e_tls_cleanup(priv);
mlx5e_ipsec_cleanup(priv);
+ mlx5e_fs_cleanup(priv);
}
static int mlx5e_init_nic_rx(struct mlx5e_priv *priv)
--
2.31.1
^ permalink raw reply related [flat|nested] 7+ messages in thread* [net 4/5] net/mlx5e: IPsec: Fix a misuse of the software parser's fields
2021-10-20 17:56 [pull request][net 0/5] mlx5 fixes 2021-10-20 Saeed Mahameed
` (2 preceding siblings ...)
2021-10-20 17:56 ` [net 3/5] net/mlx5e: Fix vlan data lost during suspend flow Saeed Mahameed
@ 2021-10-20 17:56 ` Saeed Mahameed
2021-10-20 17:56 ` [net 5/5] net/mlx5e: IPsec: Fix work queue entry ethernet segment checksum flags Saeed Mahameed
4 siblings, 0 replies; 7+ messages in thread
From: Saeed Mahameed @ 2021-10-20 17:56 UTC (permalink / raw)
To: David S. Miller, Jakub Kicinski
Cc: netdev, Emeel Hakim, Raed Salem, Saeed Mahameed
From: Emeel Hakim <ehakim@nvidia.com>
IPsec crypto offload current Software Parser (SWP) fields settings in
the ethernet segment (eseg) are not aligned with PRM/HW expectations.
Among others in case of IP|ESP|TCP packet, current driver sets the
offsets for inner_l3 and inner_l4 although there is no inner l3/l4
headers relative to ESP header in such packets.
SWP provides the offsets for HW ,so it can be used to find csum fields
to offload the checksum, however these are not necessarily used by HW
and are used as fallback in case HW fails to parse the packet, e.g
when performing IPSec Transport Aware (IP | ESP | TCP) there is no
need to add SW parse on inner packet. So in some cases packets csum
was calculated correctly , whereas in other cases it failed. The later
faced csum errors (caused by wrong packet length calculations) which
led to lots of packet drops hence the low throughput.
Fix by setting the SWP fields as expected in a IP|ESP|TCP packet.
the following describe the expected SWP offsets:
* Tunnel Mode:
* SWP: OutL3 InL3 InL4
* Pkt: MAC IP ESP IP L4
*
* Transport Mode:
* SWP: OutL3 OutL4
* Pkt: MAC IP ESP L4
*
* Tunnel(VXLAN TCP/UDP) over Transport Mode
* SWP: OutL3 InL3 InL4
* Pkt: MAC IP ESP UDP VXLAN IP L4
Fixes: f1267798c980 ("net/mlx5: Fix checksum issue of VXLAN and IPsec crypto offload")
Signed-off-by: Emeel Hakim <ehakim@nvidia.com>
Reviewed-by: Raed Salem <raeds@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
---
.../mellanox/mlx5/core/en_accel/ipsec_rxtx.c | 51 ++++++++++---------
1 file changed, 27 insertions(+), 24 deletions(-)
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_accel/ipsec_rxtx.c b/drivers/net/ethernet/mellanox/mlx5/core/en_accel/ipsec_rxtx.c
index 33de8f0092a6..fb5397324aa4 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_accel/ipsec_rxtx.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_accel/ipsec_rxtx.c
@@ -141,8 +141,7 @@ static void mlx5e_ipsec_set_swp(struct sk_buff *skb,
* Pkt: MAC IP ESP IP L4
*
* Transport Mode:
- * SWP: OutL3 InL4
- * InL3
+ * SWP: OutL3 OutL4
* Pkt: MAC IP ESP L4
*
* Tunnel(VXLAN TCP/UDP) over Transport Mode
@@ -171,31 +170,35 @@ static void mlx5e_ipsec_set_swp(struct sk_buff *skb,
return;
if (!xo->inner_ipproto) {
- eseg->swp_inner_l3_offset = skb_network_offset(skb) / 2;
- eseg->swp_inner_l4_offset = skb_inner_transport_offset(skb) / 2;
- if (skb->protocol == htons(ETH_P_IPV6))
- eseg->swp_flags |= MLX5_ETH_WQE_SWP_INNER_L3_IPV6;
- if (xo->proto == IPPROTO_UDP)
+ switch (xo->proto) {
+ case IPPROTO_UDP:
+ eseg->swp_flags |= MLX5_ETH_WQE_SWP_OUTER_L4_UDP;
+ fallthrough;
+ case IPPROTO_TCP:
+ /* IP | ESP | TCP */
+ eseg->swp_outer_l4_offset = skb_inner_transport_offset(skb) / 2;
+ break;
+ default:
+ break;
+ }
+ } else {
+ /* Tunnel(VXLAN TCP/UDP) over Transport Mode */
+ switch (xo->inner_ipproto) {
+ case IPPROTO_UDP:
eseg->swp_flags |= MLX5_ETH_WQE_SWP_INNER_L4_UDP;
- return;
- }
-
- /* Tunnel(VXLAN TCP/UDP) over Transport Mode */
- switch (xo->inner_ipproto) {
- case IPPROTO_UDP:
- eseg->swp_flags |= MLX5_ETH_WQE_SWP_INNER_L4_UDP;
- fallthrough;
- case IPPROTO_TCP:
- eseg->swp_inner_l3_offset = skb_inner_network_offset(skb) / 2;
- eseg->swp_inner_l4_offset = (skb->csum_start + skb->head - skb->data) / 2;
- if (skb->protocol == htons(ETH_P_IPV6))
- eseg->swp_flags |= MLX5_ETH_WQE_SWP_INNER_L3_IPV6;
- break;
- default:
- break;
+ fallthrough;
+ case IPPROTO_TCP:
+ eseg->swp_inner_l3_offset = skb_inner_network_offset(skb) / 2;
+ eseg->swp_inner_l4_offset =
+ (skb->csum_start + skb->head - skb->data) / 2;
+ if (skb->protocol == htons(ETH_P_IPV6))
+ eseg->swp_flags |= MLX5_ETH_WQE_SWP_INNER_L3_IPV6;
+ break;
+ default:
+ break;
+ }
}
- return;
}
void mlx5e_ipsec_set_iv_esn(struct sk_buff *skb, struct xfrm_state *x,
--
2.31.1
^ permalink raw reply related [flat|nested] 7+ messages in thread* [net 5/5] net/mlx5e: IPsec: Fix work queue entry ethernet segment checksum flags
2021-10-20 17:56 [pull request][net 0/5] mlx5 fixes 2021-10-20 Saeed Mahameed
` (3 preceding siblings ...)
2021-10-20 17:56 ` [net 4/5] net/mlx5e: IPsec: Fix a misuse of the software parser's fields Saeed Mahameed
@ 2021-10-20 17:56 ` Saeed Mahameed
4 siblings, 0 replies; 7+ messages in thread
From: Saeed Mahameed @ 2021-10-20 17:56 UTC (permalink / raw)
To: David S. Miller, Jakub Kicinski; +Cc: netdev, Emeel Hakim, Saeed Mahameed
From: Emeel Hakim <ehakim@nvidia.com>
Current Work Queue Entry (WQE) checksum (csum) flags in the ethernet
segment (eseg) in case of IPsec crypto offload datapath are not aligned
with PRM/HW expectations.
Currently the driver always sets the l3_inner_csum flag in case of IPsec
because of the wrong usage of skb->encapsulation as indicator for inner
IPsec header since skb->encapsulation is always ON for IPsec packets
since IPsec itself is an encapsulation protocol. The above forced a
failing attempts of calculating csum of non-existing segments (like in
the IP|ESP|TCP packet case which does not have an l3_inner) which led
to lots of packet drops hence the low throughput.
Fix by using xo->inner_ipproto as indicator for inner IPsec header
instead of skb->encapsulation in addition to setting the csum flags
as following:
* Tunnel Mode:
* Pkt: MAC IP ESP IP L4
* CSUM: l3_cs | l3_inner_cs | l4_inner_cs
*
* Transport Mode:
* Pkt: MAC IP ESP L4
* CSUM: l3_cs [ | l4_cs (checksum partial case)]
*
* Tunnel(VXLAN TCP/UDP) over Transport Mode
* Pkt: MAC IP ESP UDP VXLAN IP L4
* CSUM: l3_cs | l3_inner_cs | l4_inner_cs
Fixes: f1267798c980 ("net/mlx5: Fix checksum issue of VXLAN and IPsec crypto offload")
Signed-off-by: Emeel Hakim <ehakim@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
---
.../net/ethernet/mellanox/mlx5/core/en_tx.c | 20 ++++++++++---------
1 file changed, 11 insertions(+), 9 deletions(-)
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_tx.c b/drivers/net/ethernet/mellanox/mlx5/core/en_tx.c
index c63d78eda606..188994d091c5 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_tx.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_tx.c
@@ -213,19 +213,18 @@ static inline void mlx5e_insert_vlan(void *start, struct sk_buff *skb, u16 ihs)
memcpy(&vhdr->h_vlan_encapsulated_proto, skb->data + cpy1_sz, cpy2_sz);
}
-/* If packet is not IP's CHECKSUM_PARTIAL (e.g. icmd packet),
- * need to set L3 checksum flag for IPsec
- */
static void
ipsec_txwqe_build_eseg_csum(struct mlx5e_txqsq *sq, struct sk_buff *skb,
struct mlx5_wqe_eth_seg *eseg)
{
+ struct xfrm_offload *xo = xfrm_offload(skb);
+
eseg->cs_flags = MLX5_ETH_WQE_L3_CSUM;
- if (skb->encapsulation) {
- eseg->cs_flags |= MLX5_ETH_WQE_L3_INNER_CSUM;
+ if (xo->inner_ipproto) {
+ eseg->cs_flags |= MLX5_ETH_WQE_L4_INNER_CSUM | MLX5_ETH_WQE_L3_INNER_CSUM;
+ } else if (likely(skb->ip_summed == CHECKSUM_PARTIAL)) {
+ eseg->cs_flags |= MLX5_ETH_WQE_L4_CSUM;
sq->stats->csum_partial_inner++;
- } else {
- sq->stats->csum_partial++;
}
}
@@ -234,6 +233,11 @@ mlx5e_txwqe_build_eseg_csum(struct mlx5e_txqsq *sq, struct sk_buff *skb,
struct mlx5e_accel_tx_state *accel,
struct mlx5_wqe_eth_seg *eseg)
{
+ if (unlikely(mlx5e_ipsec_eseg_meta(eseg))) {
+ ipsec_txwqe_build_eseg_csum(sq, skb, eseg);
+ return;
+ }
+
if (likely(skb->ip_summed == CHECKSUM_PARTIAL)) {
eseg->cs_flags = MLX5_ETH_WQE_L3_CSUM;
if (skb->encapsulation) {
@@ -249,8 +253,6 @@ mlx5e_txwqe_build_eseg_csum(struct mlx5e_txqsq *sq, struct sk_buff *skb,
eseg->cs_flags = MLX5_ETH_WQE_L3_CSUM | MLX5_ETH_WQE_L4_CSUM;
sq->stats->csum_partial++;
#endif
- } else if (unlikely(mlx5e_ipsec_eseg_meta(eseg))) {
- ipsec_txwqe_build_eseg_csum(sq, skb, eseg);
} else
sq->stats->csum_none++;
}
--
2.31.1
^ permalink raw reply related [flat|nested] 7+ messages in thread