Skip to contentSkip to conversation

mlx5 fixes 2021-08-26

GIT PULL Applied Saeed Mahameed 2 participants 8 messages Aug 26, 2021 – Aug 27, 2021

Subsystems (4)
lore
Share & tools
lore
#0 lore raw

From: Saeed Mahameed <[email protected]>

Hi Dave and Jakub,

This series introduces some fixes to mlx5 driver. Please pull and let me know if there is any problem.

Thanks,
Saeed.

---
The following changes since commit 73367f05b25dbd064061aee780638564d15b01d1:

Merge tag 'nfsd-5.14-1' of git://linux-nfs.org/~bfields/linux (2021-08-26 13:26:40 -0700)

are available in the Git repository at:

git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linux.git tags/mlx5-fixes-2021-08-26

for you to fetch changes up to 6cc64770fb386b10a64a1fe09328396de7bb5262:

net/mlx5: DR, fix a potential use-after-free bug (2021-08-26 15:15:42 -0700)

----------------------------------------------------------------
mlx5-fixes-2021-08-26

----------------------------------------------------------------
Dima Chumak (1):
net/mlx5: Lag, fix multipath lag activation

Dmytro Linkin (1):
net/mlx5e: Use correct eswitch for stack devices with lag

Leon Romanovsky (1):
net/mlx5: Remove all auxiliary devices at the unregister event

Maor Dickman (1):
net/mlx5: E-Switch, Set vhca id valid flag when creating indir fwd group

Roi Dayan (1):
net/mlx5e: Fix possible use-after-free deleting fdb rule

Wentao_Liang (1):
net/mlx5: DR, fix a potential use-after-free bug

 drivers/net/ethernet/mellanox/mlx5/core/dev.c          |  2 +-
 .../net/ethernet/mellanox/mlx5/core/en/tc_tun_encap.c  |  4 ++--
 drivers/net/ethernet/mellanox/mlx5/core/en_tc.c        | 18 ++++++++++++++++++
 .../net/ethernet/mellanox/mlx5/core/esw/indir_table.c  |  1 +
 drivers/net/ethernet/mellanox/mlx5/core/lag.c          |  1 +
 drivers/net/ethernet/mellanox/mlx5/core/lag_mp.c       |  8 ++++++++
 drivers/net/ethernet/mellanox/mlx5/core/lag_mp.h       |  2 ++
 .../net/ethernet/mellanox/mlx5/core/steering/dr_rule.c |  2 +-
 8 files changed, 34 insertions(+), 4 deletions(-)
#1 lore raw

PATCH 1/6 net/mlx5: Lag, fix multipath lag activation

From: Dima Chumak <[email protected]>

When handling FIB_EVENT_ENTRY_REPLACE event for a new multipath route, lag activation can be missed if a stale (struct lag_mp)->mfi pointer exists, which was associated with an older multipath route that had been removed.

Normally, when a route is removed, it triggers mlx5_lag_fib_event(), which handles FIB_EVENT_ENTRY_DEL and clears mfi pointer. But, if mlx5_lag_check_prereq() condition isn't met, for example when eswitch is in legacy mode, the fib event is skipped and mfi pointer becomes stale.

Fix by resetting mfi pointer to NULL in mlx5_deactivate_lag().

Fixes 8a66e4585979 ("net/mlx5: Change ownership model for lag")
---
 drivers/net/ethernet/mellanox/mlx5/core/lag.c    | 1 +
 drivers/net/ethernet/mellanox/mlx5/core/lag_mp.c | 8 ++++++++
 drivers/net/ethernet/mellanox/mlx5/core/lag_mp.h | 2 ++
 3 files changed, 11 insertions(+)
drivers/net/ethernet/mellanox/mlx5/core/lag.c +1 -0
index 5c043c5cc403..40ef60f562b4 100644
@@ -277,6 +277,7 @@ static int mlx5_deactivate_lag(struct mlx5_lag *ldev)
 	int err;
 
 	ldev->flags &= ~MLX5_LAG_MODE_FLAGS;
+	mlx5_lag_mp_reset(ldev);
 
 	MLX5_SET(destroy_lag_in, in, opcode, MLX5_CMD_OP_DESTROY_LAG);
 	err = mlx5_cmd_exec_in(dev0, destroy_lag, in);
drivers/net/ethernet/mellanox/mlx5/core/lag_mp.c +8 -0
index c4bf8b679541..516bfc2bd797 100644
@@ -302,6 +302,14 @@ static int mlx5_lag_fib_event(struct notifier_block *nb,
 	return NOTIFY_DONE;
 }
 
+void mlx5_lag_mp_reset(struct mlx5_lag *ldev)
+{
+	/* Clear mfi, as it might become stale when a route delete event
+	 * has been missed, see mlx5_lag_fib_route_event().
+	 */
+	ldev->lag_mp.mfi = NULL;
+}
+
 int mlx5_lag_mp_init(struct mlx5_lag *ldev)
 {
 	struct lag_mp *mp = &ldev->lag_mp;
drivers/net/ethernet/mellanox/mlx5/core/lag_mp.h +2 -0
index 258ac7b2964e..729c839397a8 100644
@@ -21,11 +21,13 @@ struct lag_mp {
 
 #ifdef CONFIG_MLX5_ESWITCH
 
+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);
 
 #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) {}
 
Signature · 1 line
2.31.1
#2 lore raw

Re: [net 1/6] net/mlx5: Lag, fix multipath lag activation

This series was applied to netdev/net.git (refs/heads/master):

Hello:

This series was applied to netdev/net.git (refs/heads/master):

Quoted 8 lines · On Thu, 26 Aug 2021 15:18:05 -0700 you wrote:

From: Dima Chumak <[email protected]>

When handling FIB_EVENT_ENTRY_REPLACE event for a new multipath route, lag activation can be missed if a stale (struct lag_mp)->mfi pointer exists, which was associated with an older multipath route that had been removed.

[...]

Here is the summary with links:
- [net,1/6] net/mlx5: Lag, fix multipath lag activation

    https://git.kernel.org/netdev/net/c/2f8b6161cca5
  - [net,2/6] net/mlx5: Remove all auxiliary devices at the unregister event
    https://git.kernel.org/netdev/net/c/8e7e2e8ed0e2
  - [net,3/6] net/mlx5e: Fix possible use-after-free deleting fdb rule
    https://git.kernel.org/netdev/net/c/9a5f9cc794e1
  - [net,4/6] net/mlx5: E-Switch, Set vhca id valid flag when creating indir fwd group
    https://git.kernel.org/netdev/net/c/ca6891f9b27d
  - [net,5/6] net/mlx5e: Use correct eswitch for stack devices with lag
    https://git.kernel.org/netdev/net/c/f9d196bd632b
  - [net,6/6] net/mlx5: DR, fix a potential use-after-free bug
    https://git.kernel.org/netdev/net/c/6cc64770fb38

You are awesome, thank you!
--
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html

#3 lore raw

PATCH 2/6 net/mlx5: Remove all auxiliary devices at the unregister event

In reply to Saeed Mahameed

From: Leon Romanovsky <[email protected]>

The call to mlx5_unregister_device() means that mlx5_core driver is removed. In such scenario, we need to disregard all other flags like attach/detach and forcibly remove all auxiliary devices.

Fixes a5ae8fc9058e ("net/mlx5e: Don't create devices during unload flow")
Tested-and-Reported-by Yicong Yang <[email protected]>
---
 drivers/net/ethernet/mellanox/mlx5/core/dev.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/net/ethernet/mellanox/mlx5/core/dev.c +1 -1
index def2156e50ee..20bb37266254 100644
@@ -397,7 +397,7 @@ int mlx5_register_device(struct mlx5_core_dev *dev)
 void mlx5_unregister_device(struct mlx5_core_dev *dev)
 {
 	mutex_lock(&mlx5_intf_mutex);
-	dev->priv.flags |= MLX5_PRIV_FLAGS_DISABLE_ALL_ADEV;
+	dev->priv.flags = MLX5_PRIV_FLAGS_DISABLE_ALL_ADEV;
 	mlx5_rescan_drivers_locked(dev);
 	mutex_unlock(&mlx5_intf_mutex);
 }
Signature · 1 line
2.31.1
#4 lore raw

PATCH 3/6 net/mlx5e: Fix possible use-after-free deleting fdb rule

In reply to Saeed Mahameed

From: Roi Dayan <[email protected]>

After neigh-update-add failure we are still with a slow path rule but the driver always assume the rule is an fdb rule.
Fix neigh-update-del by checking slow path tc flag on the flow. Also fix neigh-update-add for when neigh-update-del fails the same.

Fixes 5dbe906ff1d5 ("net/mlx5e: Use a slow path rule instead if vxlan neighbour isn't available")
---
 drivers/net/ethernet/mellanox/mlx5/core/en/tc_tun_encap.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/net/ethernet/mellanox/mlx5/core/en/tc_tun_encap.c +2 -2
index 2e846b741280..1c44c6c345f5 100644
@@ -147,7 +147,7 @@ void mlx5e_tc_encap_flows_add(struct mlx5e_priv *priv,
 	mlx5e_rep_queue_neigh_stats_work(priv);
 
 	list_for_each_entry(flow, flow_list, tmp_list) {
-		if (!mlx5e_is_offloaded_flow(flow))
+		if (!mlx5e_is_offloaded_flow(flow) || !flow_flag_test(flow, SLOW))
 			continue;
 		attr = flow->attr;
 		esw_attr = attr->esw_attr;
@@ -188,7 +188,7 @@ void mlx5e_tc_encap_flows_del(struct mlx5e_priv *priv,
 	int err;
 
 	list_for_each_entry(flow, flow_list, tmp_list) {
-		if (!mlx5e_is_offloaded_flow(flow))
+		if (!mlx5e_is_offloaded_flow(flow) || flow_flag_test(flow, SLOW))
 			continue;
 		attr = flow->attr;
 		esw_attr = attr->esw_attr;
Signature · 1 line
2.31.1
#5 lore raw

PATCH 4/6 net/mlx5: E-Switch, Set vhca id valid flag when creating indir fwd group

In reply to Saeed Mahameed

From: Maor Dickman <[email protected]>

When indirect forward group is created, flow is added with vhca id but without setting vhca id valid flag which violates the PRM.

Fix by setting the missing flag, vhca id valid.

Fixes 34ca65352ddf ("net/mlx5: E-Switch, Indirect table infrastructure")
---
 drivers/net/ethernet/mellanox/mlx5/core/esw/indir_table.c | 1 +
 1 file changed, 1 insertion(+)
drivers/net/ethernet/mellanox/mlx5/core/esw/indir_table.c +1 -0
index 3da7becc1069..425c91814b34 100644
@@ -364,6 +364,7 @@ static int mlx5_create_indir_fwd_group(struct mlx5_eswitch *esw,
 	dest.type = MLX5_FLOW_DESTINATION_TYPE_VPORT;
 	dest.vport.num = e->vport;
 	dest.vport.vhca_id = MLX5_CAP_GEN(esw->dev, vhca_id);
+	dest.vport.flags = MLX5_FLOW_DEST_VPORT_VHCA_ID;
 	e->fwd_rule = mlx5_add_flow_rules(e->ft, spec, &flow_act, &dest, 1);
 	if (IS_ERR(e->fwd_rule)) {
 		mlx5_destroy_flow_group(e->fwd_grp);
Signature · 1 line
2.31.1
#6 lore raw

PATCH 5/6 net/mlx5e: Use correct eswitch for stack devices with lag

In reply to Saeed Mahameed

From: Dmytro Linkin <[email protected]>

If link aggregation is used within stack devices driver rejects encap rules if PF of the VF tunnel device is down. This happens because route resolved for other PF and its eswitch instance is used to determine correct vport.
To fix that use devcom feature to retrieve other eswitch instance if failed to find vport for the 1st eswitch and LAG is active.

Fixes 10742efc20a4 ("net/mlx5e: VF tunnel TX traffic offloading")
---
 .../net/ethernet/mellanox/mlx5/core/en_tc.c    | 18 ++++++++++++++++++
 1 file changed, 18 insertions(+)
drivers/net/ethernet/mellanox/mlx5/core/en_tc.c +18 -0
index d273758255c3..6eba574c5a36 100644
@@ -1338,6 +1338,7 @@ bool mlx5e_tc_is_vf_tunnel(struct net_device *out_dev, struct net_device *route_
 int mlx5e_tc_query_route_vport(struct net_device *out_dev, struct net_device *route_dev, u16 *vport)
 {
 	struct mlx5e_priv *out_priv, *route_priv;
+	struct mlx5_devcom *devcom = NULL;
 	struct mlx5_core_dev *route_mdev;
 	struct mlx5_eswitch *esw;
 	u16 vhca_id;
@@ -1349,7 +1350,24 @@ int mlx5e_tc_query_route_vport(struct net_device *out_dev, struct net_device *ro
 	route_mdev = route_priv->mdev;
 
 	vhca_id = MLX5_CAP_GEN(route_mdev, vhca_id);
+	if (mlx5_lag_is_active(out_priv->mdev)) {
+		/* In lag case we may get devices from different eswitch instances.
+		 * If we failed to get vport num, it means, mostly, that we on the wrong
+		 * eswitch.
+		 */
+		err = mlx5_eswitch_vhca_id_to_vport(esw, vhca_id, vport);
+		if (err != -ENOENT)
+			return err;
+
+		devcom = out_priv->mdev->priv.devcom;
+		esw = mlx5_devcom_get_peer_data(devcom, MLX5_DEVCOM_ESW_OFFLOADS);
+		if (!esw)
+			return -ENODEV;
+	}
+
 	err = mlx5_eswitch_vhca_id_to_vport(esw, vhca_id, vport);
+	if (devcom)
+		mlx5_devcom_release_peer_data(devcom, MLX5_DEVCOM_ESW_OFFLOADS);
 	return err;
 }
 
Signature · 1 line
2.31.1
#7 lore raw

PATCH 6/6 net/mlx5: DR, fix a potential use-after-free bug

In reply to Saeed Mahameed

From: Wentao_Liang <[email protected]>

In line 849 (#1), "mlx5dr_htbl_put(cur_htbl);" drops the reference to cur_htbl and may cause cur_htbl to be freed.

However, cur_htbl is subsequently used in the next line, which may result in an use-after-free bug.

Fix this by calling mlx5dr_err() before the cur_htbl is put.

---
 drivers/net/ethernet/mellanox/mlx5/core/steering/dr_rule.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/net/ethernet/mellanox/mlx5/core/steering/dr_rule.c +1 -1
index 43356fad53de..ffdfb5a94b14 100644
@@ -846,9 +846,9 @@ dr_rule_handle_ste_branch(struct mlx5dr_rule *rule,
 			new_htbl = dr_rule_rehash(rule, nic_rule, cur_htbl,
 						  ste_location, send_ste_list);
 			if (!new_htbl) {
-				mlx5dr_htbl_put(cur_htbl);
 				mlx5dr_err(dmn, "Failed creating rehash table, htbl-log_size: %d\n",
 					   cur_htbl->chunk_size);
+				mlx5dr_htbl_put(cur_htbl);
 			} else {
 				cur_htbl = new_htbl;
 			}
Signature · 1 line
2.31.1