Skip to contentSkip to conversation

mlx5 fixes 2022-11-02

GIT PULL Superseded Saeed Mahameed 3 participants 15 messages Nov 5, 2022 – Nov 8, 2022

Subsystems (5)

v2 of 3 Superseded Latest is v3 →

lore
Share & tools
lore
#0 lore raw

From: Saeed Mahameed <[email protected]>

v1->v2:
- Fix 32bit build issue.

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

Thanks,
Saeed.

The following changes since commit b7cbc6740bd6ad5d43345a2504f7e4beff0d709f:

net: fman: Unregister ethernet device on removal (2022-11-04 19:33:38 -0700)

are available in the Git repository at:

git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linux.git tags/mlx5-fixes-2022-11-02

for you to fetch changes up to 4a1ed426f330a544f8b56c2e4fae6a8c4fa77ccd:

net/mlx5e: TC, Fix slab-out-of-bounds in parse_tc_actions (2022-11-05 00:04:34 -0700)

----------------------------------------------------------------
mlx5-fixes-2022-11-02

----------------------------------------------------------------
Chris Mi (1):
net/mlx5: E-switch, Set to legacy mode if failed to change switchdev mode

Jianbo Liu (1):
net/mlx5e: TC, Fix wrong rejection of packet-per-second policing

Maxim Mikityanskiy (2):
net/mlx5e: Add missing sanity checks for max TX WQE size net/mlx5e: Fix usage of DMA sync API

Moshe Shemesh (1):
net/mlx5: Fix possible deadlock on mlx5e_tx_timeout_work

Roi Dayan (3):
net/mlx5e: Fix tc acts array not to be dependent on enum order net/mlx5e: E-Switch, Fix comparing termination table instance net/mlx5e: TC, Fix slab-out-of-bounds in parse_tc_actions

Roy Novich (1):
net/mlx5: Allow async trigger completion execution on single CPU systems

Shay Drory (1):
net/mlx5: fw_reset: Don't try to load device in case PCI isn't working

Vlad Buslov (1):
net/mlx5: Bridge, verify LAG state when adding bond to bridge

 drivers/net/ethernet/mellanox/mlx5/core/cmd.c      | 11 ++-
 .../ethernet/mellanox/mlx5/core/en/rep/bridge.c    | 31 ++++++++
 .../ethernet/mellanox/mlx5/core/en/tc/act/act.c    | 92 ++++++++--------------
 drivers/net/ethernet/mellanox/mlx5/core/en/txrx.h  | 24 +++++-
 drivers/net/ethernet/mellanox/mlx5/core/en/xdp.c   |  4 +-
 drivers/net/ethernet/mellanox/mlx5/core/en_main.c  | 13 ++-
 drivers/net/ethernet/mellanox/mlx5/core/en_rx.c    | 27 ++++---
 drivers/net/ethernet/mellanox/mlx5/core/en_tc.c    | 14 ++--
 drivers/net/ethernet/mellanox/mlx5/core/en_tx.c    |  5 ++
 drivers/net/ethernet/mellanox/mlx5/core/eswitch.c  | 14 ++--
 .../ethernet/mellanox/mlx5/core/eswitch_offloads.c | 18 +----
 .../mellanox/mlx5/core/eswitch_offloads_termtbl.c  | 14 ++--
 drivers/net/ethernet/mellanox/mlx5/core/fw_reset.c |  3 +-
 13 files changed, 153 insertions(+), 117 deletions(-)
#1 lore raw

PATCH v2 1/11 net/mlx5: Bridge, verify LAG state when adding bond to bridge

From: Vlad Buslov <[email protected]>

Mlx5 LAG is initialized asynchronously on a workqueue which means that for a brief moment after setting mlx5 UL representors as lower devices of a bond netdevice the LAG itself is not fully initialized in the driver. When adding such bond device to a bridge mlx5 bridge code will not consider it as offload-capable, skip creating necessary bookkeeping and fail any further bridge offload-related commands with it (setting VLANs, offloading FDBs, etc.). In order to make the error explicit during bridge initialization stage implement the code that detects such condition during NETDEV_PRECHANGEUPPER event and returns an error.

Fixes ff9b7521468b ("net/mlx5: Bridge, support LAG")
---
 .../mellanox/mlx5/core/en/rep/bridge.c        | 31 +++++++++++++++++++
 1 file changed, 31 insertions(+)
drivers/net/ethernet/mellanox/mlx5/core/en/rep/bridge.c +31 -0
index 39ef2a2561a3..8099a21e674c 100644
@@ -164,6 +164,36 @@ static int mlx5_esw_bridge_port_changeupper(struct notifier_block *nb, void *ptr
 	return err;
 }
 
+static int
+mlx5_esw_bridge_changeupper_validate_netdev(void *ptr)
+{
+	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
+	struct netdev_notifier_changeupper_info *info = ptr;
+	struct net_device *upper = info->upper_dev;
+	struct net_device *lower;
+	struct list_head *iter;
+
+	if (!netif_is_bridge_master(upper) || !netif_is_lag_master(dev))
+		return 0;
+
+	netdev_for_each_lower_dev(dev, lower, iter) {
+		struct mlx5_core_dev *mdev;
+		struct mlx5e_priv *priv;
+
+		if (!mlx5e_eswitch_rep(lower))
+			continue;
+
+		priv = netdev_priv(lower);
+		mdev = priv->mdev;
+		if (!mlx5_lag_is_active(mdev))
+			return -EAGAIN;
+		if (!mlx5_lag_is_shared_fdb(mdev))
+			return -EOPNOTSUPP;
+	}
+
+	return 0;
+}
+
 static int mlx5_esw_bridge_switchdev_port_event(struct notifier_block *nb,
 						unsigned long event, void *ptr)
 {
@@ -171,6 +201,7 @@ static int mlx5_esw_bridge_switchdev_port_event(struct notifier_block *nb,
 
 	switch (event) {
 	case NETDEV_PRECHANGEUPPER:
+		err = mlx5_esw_bridge_changeupper_validate_netdev(ptr);
 		break;
 
 	case NETDEV_CHANGEUPPER:
Signature · 1 line
2.38.1
#3 lore raw

PATCH v2 3/11 net/mlx5: E-switch, Set to legacy mode if failed to change switchdev mode

In reply to Saeed Mahameed

From: Chris Mi <[email protected]>

No need to rollback to the other mode because probably will fail again. Just set to legacy mode and clear fdb table created flag. So that fdb table will not be cleared again.

Fixes f019679ea5f2 ("net/mlx5: E-switch, Remove dependency between sriov and eswitch mode")
---
 .../net/ethernet/mellanox/mlx5/core/eswitch.c  | 14 ++++++++------
 .../mellanox/mlx5/core/eswitch_offloads.c      | 18 +++---------------
 2 files changed, 11 insertions(+), 21 deletions(-)
drivers/net/ethernet/mellanox/mlx5/core/eswitch.c +8 -6
index c59107fa9e6d..2169486c4bfb 100644
@@ -1387,12 +1387,14 @@ void mlx5_eswitch_disable_locked(struct mlx5_eswitch *esw)
 		 esw->mode == MLX5_ESWITCH_LEGACY ? "LEGACY" : "OFFLOADS",
 		 esw->esw_funcs.num_vfs, esw->enabled_vports);
 
-	esw->fdb_table.flags &= ~MLX5_ESW_FDB_CREATED;
-	if (esw->mode == MLX5_ESWITCH_OFFLOADS)
-		esw_offloads_disable(esw);
-	else if (esw->mode == MLX5_ESWITCH_LEGACY)
-		esw_legacy_disable(esw);
-	mlx5_esw_acls_ns_cleanup(esw);
+	if (esw->fdb_table.flags & MLX5_ESW_FDB_CREATED) {
+		esw->fdb_table.flags &= ~MLX5_ESW_FDB_CREATED;
+		if (esw->mode == MLX5_ESWITCH_OFFLOADS)
+			esw_offloads_disable(esw);
+		else if (esw->mode == MLX5_ESWITCH_LEGACY)
+			esw_legacy_disable(esw);
+		mlx5_esw_acls_ns_cleanup(esw);
+	}
 
 	if (esw->mode == MLX5_ESWITCH_OFFLOADS)
 		devl_rate_nodes_destroy(devlink);
drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c +3 -15
index 4e50df3139c6..728ca9f2bb9d 100644
@@ -2310,7 +2310,7 @@ static int esw_create_restore_table(struct mlx5_eswitch *esw)
 static int esw_offloads_start(struct mlx5_eswitch *esw,
 			      struct netlink_ext_ack *extack)
 {
-	int err, err1;
+	int err;
 
 	esw->mode = MLX5_ESWITCH_OFFLOADS;
 	err = mlx5_eswitch_enable_locked(esw, esw->dev->priv.sriov.num_vfs);
@@ -2318,11 +2318,6 @@ static int esw_offloads_start(struct mlx5_eswitch *esw,
 		NL_SET_ERR_MSG_MOD(extack,
 				   "Failed setting eswitch to offloads");
 		esw->mode = MLX5_ESWITCH_LEGACY;
-		err1 = mlx5_eswitch_enable_locked(esw, MLX5_ESWITCH_IGNORE_NUM_VFS);
-		if (err1) {
-			NL_SET_ERR_MSG_MOD(extack,
-					   "Failed setting eswitch back to legacy");
-		}
 		mlx5_rescan_drivers(esw->dev);
 	}
 	if (esw->offloads.inline_mode == MLX5_INLINE_MODE_NONE) {
@@ -3389,19 +3384,12 @@ int esw_offloads_enable(struct mlx5_eswitch *esw)
 static int esw_offloads_stop(struct mlx5_eswitch *esw,
 			     struct netlink_ext_ack *extack)
 {
-	int err, err1;
+	int err;
 
 	esw->mode = MLX5_ESWITCH_LEGACY;
 	err = mlx5_eswitch_enable_locked(esw, MLX5_ESWITCH_IGNORE_NUM_VFS);
-	if (err) {
+	if (err)
 		NL_SET_ERR_MSG_MOD(extack, "Failed setting eswitch to legacy");
-		esw->mode = MLX5_ESWITCH_OFFLOADS;
-		err1 = mlx5_eswitch_enable_locked(esw, MLX5_ESWITCH_IGNORE_NUM_VFS);
-		if (err1) {
-			NL_SET_ERR_MSG_MOD(extack,
-					   "Failed setting eswitch back to offloads");
-		}
-	}
 
 	return err;
 }
Signature · 1 line
2.38.1
#5 lore raw

PATCH v2 5/11 net/mlx5: Fix possible deadlock on mlx5e_tx_timeout_work

In reply to Saeed Mahameed

From: Moshe Shemesh <[email protected]>

Due to the cited patch, devlink health commands take devlink lock and this may result in deadlock for mlx5e_tx_reporter as it takes local state_lock before calling devlink health report and on the other hand devlink health commands such as diagnose for same reporter take local state_lock after taking devlink lock (see kernel log below).

To fix it, remove local state_lock from mlx5e_tx_timeout_work() before calling devlink_health_report() and take care to cancel the work before any call to close channels, which may free the SQs that should be handled by the work. Before cancel_work_sync(), use current_work() to check we are not calling it from within the work, as mlx5e_tx_timeout_work() itself may close the channels and reopen as part of recovery flow.

Kernel log:
======================================================
WARNING: possible circular locking dependency detected 6.0.0-rc3_for_upstream_debug_2022_08_30_13_10 #1 Not tainted
------------------------------------------------------
kworker/u16:2/65 is trying to acquire lock:
ffff888122f6c2f8 (&devlink->lock_key#2){+.+.}-{3:3}, at: devlink_health_report+0x2f1/0x7e0

but task is already holding lock:
ffff888121d20be0 (&priv->state_lock){+.+.}-{3:3}, at: mlx5e_tx_timeout_work+0x70/0x280 [mlx5_core]

which lock already depends on the new lock.

the existing dependency chain (in reverse order) is:

-> #1 (&priv->state_lock){+.+.}-{3:3}:

       __mutex_lock+0x12c/0x14b0
       mlx5e_rx_reporter_diagnose+0x71/0x700 [mlx5_core]
       devlink_nl_cmd_health_reporter_diagnose_doit+0x212/0xa50
       genl_family_rcv_msg_doit+0x1e9/0x2f0
       genl_rcv_msg+0x2e9/0x530
       netlink_rcv_skb+0x11d/0x340
       genl_rcv+0x24/0x40
       netlink_unicast+0x438/0x710
       netlink_sendmsg+0x788/0xc40
       sock_sendmsg+0xb0/0xe0
       __sys_sendto+0x1c1/0x290
       __x64_sys_sendto+0xdd/0x1b0
       do_syscall_64+0x3d/0x90
       entry_SYSCALL_64_after_hwframe+0x46/0xb0

-> #0 (&devlink->lock_key#2){+.+.}-{3:3}:

       __lock_acquire+0x2c8a/0x6200
       lock_acquire+0x1c1/0x550
       __mutex_lock+0x12c/0x14b0
       devlink_health_report+0x2f1/0x7e0
       mlx5e_health_report+0xc9/0xd7 [mlx5_core]
       mlx5e_reporter_tx_timeout+0x2ab/0x3d0 [mlx5_core]
       mlx5e_tx_timeout_work+0x1c1/0x280 [mlx5_core]
       process_one_work+0x7c2/0x1340
       worker_thread+0x59d/0xec0
       kthread+0x28f/0x330
       ret_from_fork+0x1f/0x30

other info that might help us debug this:

 Possible unsafe locking scenario:
       CPU0                    CPU1
       ----                    ----
  lock(&priv->state_lock);
                               lock(&devlink->lock_key#2);
                               lock(&priv->state_lock);

lock(&devlink->lock_key#2);

*** DEADLOCK ***

4 locks held by kworker/u16:2/65:
#0: ffff88811a55b138 ((wq_completion)mlx5e#2){+.+.}-{0:0}, at: process_one_work+0x6e2/0x1340 #1: ffff888101de7db8 ((work_completion)(&priv->tx_timeout_work)){+.+.}-{0:0}, at: process_one_work+0x70f/0x1340 #2: ffffffff84ce8328 (rtnl_mutex){+.+.}-{3:3}, at: mlx5e_tx_timeout_work+0x53/0x280 [mlx5_core] #3: ffff888121d20be0 (&priv->state_lock){+.+.}-{3:3}, at: mlx5e_tx_timeout_work+0x70/0x280 [mlx5_core]

stack backtrace:
CPU: 1 PID: 65 Comm: kworker/u16:2 Not tainted 6.0.0-rc3_for_upstream_debug_2022_08_30_13_10 #1
Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.16.0-0-gd239552ce722-prebuilt.qemu.org 04/01/2014
Workqueue: mlx5e mlx5e_tx_timeout_work [mlx5_core]
Call Trace:
 <TASK>
 dump_stack_lvl+0x57/0x7d
 check_noncircular+0x278/0x300
 ? print_circular_bug+0x460/0x460
 ? find_held_lock+0x2d/0x110
 ? __stack_depot_save+0x24c/0x520
 ? alloc_chain_hlocks+0x228/0x700
 __lock_acquire+0x2c8a/0x6200
 ? register_lock_class+0x1860/0x1860
 ? kasan_save_stack+0x1e/0x40
 ? kasan_set_free_info+0x20/0x30
 ? ____kasan_slab_free+0x11d/0x1b0
 ? kfree+0x1ba/0x520
 ? devlink_health_do_dump.part.0+0x171/0x3a0
 ? devlink_health_report+0x3d5/0x7e0
 lock_acquire+0x1c1/0x550
 ? devlink_health_report+0x2f1/0x7e0
 ? lockdep_hardirqs_on_prepare+0x400/0x400
 ? find_held_lock+0x2d/0x110
 __mutex_lock+0x12c/0x14b0
 ? devlink_health_report+0x2f1/0x7e0
 ? devlink_health_report+0x2f1/0x7e0
 ? mutex_lock_io_nested+0x1320/0x1320
 ? trace_hardirqs_on+0x2d/0x100
 ? bit_wait_io_timeout+0x170/0x170
 ? devlink_health_do_dump.part.0+0x171/0x3a0
 ? kfree+0x1ba/0x520
 ? devlink_health_do_dump.part.0+0x171/0x3a0
 devlink_health_report+0x2f1/0x7e0
 mlx5e_health_report+0xc9/0xd7 [mlx5_core]
 mlx5e_reporter_tx_timeout+0x2ab/0x3d0 [mlx5_core]
 ? lockdep_hardirqs_on_prepare+0x400/0x400
 ? mlx5e_reporter_tx_err_cqe+0x1b0/0x1b0 [mlx5_core]
 ? mlx5e_tx_reporter_timeout_dump+0x70/0x70 [mlx5_core]
 ? mlx5e_tx_reporter_dump_sq+0x320/0x320 [mlx5_core]
 ? mlx5e_tx_timeout_work+0x70/0x280 [mlx5_core]
 ? mutex_lock_io_nested+0x1320/0x1320
 ? process_one_work+0x70f/0x1340
 ? lockdep_hardirqs_on_prepare+0x400/0x400
 ? lock_downgrade+0x6e0/0x6e0
 mlx5e_tx_timeout_work+0x1c1/0x280 [mlx5_core]
 process_one_work+0x7c2/0x1340
 ? lockdep_hardirqs_on_prepare+0x400/0x400
 ? pwq_dec_nr_in_flight+0x230/0x230
 ? rwlock_bug.part.0+0x90/0x90
 worker_thread+0x59d/0xec0
 ? process_one_work+0x1340/0x1340
 kthread+0x28f/0x330
 ? kthread_complete_and_exit+0x20/0x20
 ret_from_fork+0x1f/0x30
 </TASK>
Fixes c90005b5f75c ("devlink: Hold the instance lock in health callbacks")
---
 drivers/net/ethernet/mellanox/mlx5/core/en_main.c | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)
drivers/net/ethernet/mellanox/mlx5/core/en_main.c +4 -2
index 364f04309149..09296f1ae448 100644
@@ -2954,6 +2954,9 @@ static int mlx5e_switch_priv_channels(struct mlx5e_priv *priv,
 	netif_carrier_off(netdev);
 
 	mlx5e_deactivate_priv_channels(priv);
+	/* Once deactivated, new tx_timeout_work won't be initiated. */
+	if (current_work() != &priv->tx_timeout_work)
+		cancel_work_sync(&priv->tx_timeout_work);
 
 	old_chs = priv->channels;
 	priv->channels = *new_chs;
@@ -3106,6 +3109,7 @@ int mlx5e_close_locked(struct net_device *netdev)
 
 	netif_carrier_off(priv->netdev);
 	mlx5e_deactivate_priv_channels(priv);
+	cancel_work_sync(&priv->tx_timeout_work);
 	mlx5e_close_channels(&priv->channels);
 
 	return 0;
@@ -4657,7 +4661,6 @@ static void mlx5e_tx_timeout_work(struct work_struct *work)
 	int i;
 
 	rtnl_lock();
-	mutex_lock(&priv->state_lock);
 
 	if (!test_bit(MLX5E_STATE_OPENED, &priv->state))
 		goto unlock;
@@ -4676,7 +4679,6 @@ static void mlx5e_tx_timeout_work(struct work_struct *work)
 	}
 
 unlock:
-	mutex_unlock(&priv->state_lock);
 	rtnl_unlock();
 }
 
Signature · 1 line
2.38.1
#4 lore raw

PATCH v2 4/11 net/mlx5: fw_reset: Don't try to load device in case PCI isn't working

In reply to Saeed Mahameed

From: Shay Drory <[email protected]>

In case PCI reads fail after unload, there is no use in trying to load the device.

Fixes 5ec697446f46 ("net/mlx5: Add support for devlink reload action fw activate")
---
 drivers/net/ethernet/mellanox/mlx5/core/fw_reset.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
drivers/net/ethernet/mellanox/mlx5/core/fw_reset.c +2 -1
index 07c583996c29..9d908a0ccfef 100644
@@ -152,7 +152,8 @@ static void mlx5_fw_reset_complete_reload(struct mlx5_core_dev *dev)
 		mlx5_unload_one(dev);
 		if (mlx5_health_wait_pci_up(dev))
 			mlx5_core_err(dev, "reset reload flow aborted, PCI reads still not working\n");
-		mlx5_load_one(dev, false);
+		else
+			mlx5_load_one(dev, false);
 		devlink_remote_reload_actions_performed(priv_to_devlink(dev), 0,
 							BIT(DEVLINK_RELOAD_ACTION_DRIVER_REINIT) |
 							BIT(DEVLINK_RELOAD_ACTION_FW_ACTIVATE));
Signature · 1 line
2.38.1
#9 lore raw

PATCH v2 6/11 net/mlx5e: Add missing sanity checks for max TX WQE size

In reply to Saeed Mahameed

From: Maxim Mikityanskiy <[email protected]>

The commit cited below started using the firmware capability for the maximum TX WQE size. This commit adds an important check to verify that the driver doesn't attempt to exceed this capability, and also restores another check mistakenly removed in the cited commit (a WQE must not exceed the page size).

Fixes c27bd1718c06 ("net/mlx5e: Read max WQEBBs on the SQ from firmware")
---
 .../net/ethernet/mellanox/mlx5/core/en/txrx.h | 24 ++++++++++++++++++-
 .../net/ethernet/mellanox/mlx5/core/en_main.c |  7 ++++++
 .../net/ethernet/mellanox/mlx5/core/en_tx.c   |  5 ++++
 3 files changed, 35 insertions(+), 1 deletion(-)
drivers/net/ethernet/mellanox/mlx5/core/en/txrx.h +23 -1
index cb164b62f543..853f312cd757 100644
@@ -11,6 +11,27 @@
 
 #define INL_HDR_START_SZ (sizeof(((struct mlx5_wqe_eth_seg *)NULL)->inline_hdr.start))
 
+/* IPSEC inline data includes:
+ * 1. ESP trailer: up to 255 bytes of padding, 1 byte for pad length, 1 byte for
+ *    next header.
+ * 2. ESP authentication data: 16 bytes for ICV.
+ */
+#define MLX5E_MAX_TX_IPSEC_DS DIV_ROUND_UP(sizeof(struct mlx5_wqe_inline_seg) + \
+					   255 + 1 + 1 + 16, MLX5_SEND_WQE_DS)
+
+/* 366 should be big enough to cover all L2, L3 and L4 headers with possible
+ * encapsulations.
+ */
+#define MLX5E_MAX_TX_INLINE_DS DIV_ROUND_UP(366 - INL_HDR_START_SZ + VLAN_HLEN, \
+					    MLX5_SEND_WQE_DS)
+
+/* Sync the calculation with mlx5e_sq_calc_wqe_attr. */
+#define MLX5E_MAX_TX_WQEBBS DIV_ROUND_UP(MLX5E_TX_WQE_EMPTY_DS_COUNT + \
+					 MLX5E_MAX_TX_INLINE_DS + \
+					 MLX5E_MAX_TX_IPSEC_DS + \
+					 MAX_SKB_FRAGS + 1, \
+					 MLX5_SEND_WQEBB_NUM_DS)
+
 #define MLX5E_RX_ERR_CQE(cqe) (get_cqe_opcode(cqe) != MLX5_CQE_RESP_SEND)
 
 static inline
@@ -424,6 +445,8 @@ mlx5e_set_eseg_swp(struct sk_buff *skb, struct mlx5_wqe_eth_seg *eseg,
 
 static inline u16 mlx5e_stop_room_for_wqe(struct mlx5_core_dev *mdev, u16 wqe_size)
 {
+	WARN_ON_ONCE(PAGE_SIZE / MLX5_SEND_WQE_BB < mlx5e_get_max_sq_wqebbs(mdev));
+
 	/* A WQE must not cross the page boundary, hence two conditions:
 	 * 1. Its size must not exceed the page size.
 	 * 2. If the WQE size is X, and the space remaining in a page is less
@@ -436,7 +459,6 @@ static inline u16 mlx5e_stop_room_for_wqe(struct mlx5_core_dev *mdev, u16 wqe_si
 		  "wqe_size %u is greater than max SQ WQEBBs %u",
 		  wqe_size, mlx5e_get_max_sq_wqebbs(mdev));
 
-
 	return MLX5E_STOP_ROOM(wqe_size);
 }
 
drivers/net/ethernet/mellanox/mlx5/core/en_main.c +7 -0
index 09296f1ae448..76861279fbda 100644
@@ -5696,6 +5696,13 @@ int mlx5e_attach_netdev(struct mlx5e_priv *priv)
 		mlx5e_fs_set_state_destroy(priv->fs,
 					   !test_bit(MLX5E_STATE_DESTROYING, &priv->state));
 
+	/* Validate the max_wqe_size_sq capability. */
+	if (WARN_ON_ONCE(mlx5e_get_max_sq_wqebbs(priv->mdev) < MLX5E_MAX_TX_WQEBBS)) {
+		mlx5_core_warn(priv->mdev, "MLX5E: Max SQ WQEBBs firmware capability: %u, needed %lu\n",
+			       mlx5e_get_max_sq_wqebbs(priv->mdev), MLX5E_MAX_TX_WQEBBS);
+		return -EIO;
+	}
+
 	/* max number of channels may have changed */
 	max_nch = mlx5e_calc_max_nch(priv->mdev, priv->netdev, profile);
 	if (priv->channels.params.num_channels > max_nch) {
drivers/net/ethernet/mellanox/mlx5/core/en_tx.c +5 -0
index 6adca01fbdc9..f7897ddb29c5 100644
@@ -305,6 +305,8 @@ static void mlx5e_sq_calc_wqe_attr(struct sk_buff *skb, const struct mlx5e_tx_at
 	u16 ds_cnt_inl = 0;
 	u16 ds_cnt_ids = 0;
 
+	/* Sync the calculation with MLX5E_MAX_TX_WQEBBS. */
+
 	if (attr->insz)
 		ds_cnt_ids = DIV_ROUND_UP(sizeof(struct mlx5_wqe_inline_seg) + attr->insz,
 					  MLX5_SEND_WQE_DS);
@@ -317,6 +319,9 @@ static void mlx5e_sq_calc_wqe_attr(struct sk_buff *skb, const struct mlx5e_tx_at
 			inl += VLAN_HLEN;
 
 		ds_cnt_inl = DIV_ROUND_UP(inl, MLX5_SEND_WQE_DS);
+		if (WARN_ON_ONCE(ds_cnt_inl > MLX5E_MAX_TX_INLINE_DS))
+			netdev_warn(skb->dev, "ds_cnt_inl = %u > max %u\n", ds_cnt_inl,
+				    (u16)MLX5E_MAX_TX_INLINE_DS);
 		ds_cnt += ds_cnt_inl;
 	}
 
Signature · 1 line
2.38.1
#2 lore raw

PATCH v2 2/11 net/mlx5: Allow async trigger completion execution on single CPU systems

In reply to Saeed Mahameed

From: Roy Novich <[email protected]>

For a single CPU system, the kernel thread executing mlx5_cmd_flush() never releases the CPU but calls down_trylock(&cmd→sem) in a busy loop. On a single processor system, this leads to a deadlock as the kernel thread which executes mlx5_cmd_invoke() never gets scheduled. Fix this, by adding the cond_resched() call to the loop, allow the command completion kernel thread to execute.

Fixes 8e715cd613a1 ("net/mlx5: Set command entry semaphore up once got index free")
---
 drivers/net/ethernet/mellanox/mlx5/core/cmd.c | 11 ++++++++---
 1 file changed, 8 insertions(+), 3 deletions(-)
drivers/net/ethernet/mellanox/mlx5/core/cmd.c +8 -3
index 46ba4c2faad2..2e0d59ca62b5 100644
@@ -1770,12 +1770,17 @@ void mlx5_cmd_flush(struct mlx5_core_dev *dev)
 	struct mlx5_cmd *cmd = &dev->cmd;
 	int i;
 
-	for (i = 0; i < cmd->max_reg_cmds; i++)
-		while (down_trylock(&cmd->sem))
+	for (i = 0; i < cmd->max_reg_cmds; i++) {
+		while (down_trylock(&cmd->sem)) {
 			mlx5_cmd_trigger_completions(dev);
+			cond_resched();
+		}
+	}
 
-	while (down_trylock(&cmd->pages_sem))
+	while (down_trylock(&cmd->pages_sem)) {
 		mlx5_cmd_trigger_completions(dev);
+		cond_resched();
+	}
 
 	/* Unlock cmdif */
 	up(&cmd->pages_sem);
Signature · 1 line
2.38.1
#10 lore raw

PATCH v2 7/11 net/mlx5e: Fix usage of DMA sync API

In reply to Saeed Mahameed

From: Maxim Mikityanskiy <[email protected]>

DMA sync functions should use the same direction that was used by DMA mapping. Use DMA_BIDIRECTIONAL for XDP_TX from regular RQ, which reuses the same mapping that was used for RX, and DMA_TO_DEVICE for XDP_TX from XSK RQ and XDP_REDIRECT, which establish a new mapping in this direction. On the RX side, use the same direction that was used when setting up the mapping (DMA_BIDIRECTIONAL for XDP, DMA_FROM_DEVICE otherwise).

Also don't skip sync for device when establishing a DMA_FROM_DEVICE mapping for RX, as some architectures (ARM) may require invalidating caches before the device can use the mapping. It doesn't break the bugfix made in
commit 0b7cfa4082fb ("net/mlx5e: Fix page DMA map/unmap attributes"), since the bug happened on unmap.

Fixes 0b7cfa4082fb ("net/mlx5e: Fix page DMA map/unmap attributes")
Fixes b5503b994ed5 ("net/mlx5e: XDP TX forwarding support")
---
 .../net/ethernet/mellanox/mlx5/core/en/xdp.c  |  4 +--
 .../net/ethernet/mellanox/mlx5/core/en_rx.c   | 27 ++++++++++---------
 2 files changed, 16 insertions(+), 15 deletions(-)
drivers/net/ethernet/mellanox/mlx5/core/en/xdp.c +2 -2
index 4685c652c97e..20507ef2f956 100644
@@ -117,7 +117,7 @@ mlx5e_xmit_xdp_buff(struct mlx5e_xdpsq *sq, struct mlx5e_rq *rq,
 	xdpi.page.rq = rq;
 
 	dma_addr = page_pool_get_dma_addr(page) + (xdpf->data - (void *)xdpf);
-	dma_sync_single_for_device(sq->pdev, dma_addr, xdptxd.len, DMA_TO_DEVICE);
+	dma_sync_single_for_device(sq->pdev, dma_addr, xdptxd.len, DMA_BIDIRECTIONAL);
 
 	if (unlikely(xdp_frame_has_frags(xdpf))) {
 		sinfo = xdp_get_shared_info_from_frame(xdpf);
@@ -131,7 +131,7 @@ mlx5e_xmit_xdp_buff(struct mlx5e_xdpsq *sq, struct mlx5e_rq *rq,
 				skb_frag_off(frag);
 			len = skb_frag_size(frag);
 			dma_sync_single_for_device(sq->pdev, addr, len,
-						   DMA_TO_DEVICE);
+						   DMA_BIDIRECTIONAL);
 		}
 	}
 
drivers/net/ethernet/mellanox/mlx5/core/en_rx.c +14 -13
index 58084650151f..a61a43fc8d5c 100644
@@ -266,7 +266,7 @@ static inline bool mlx5e_rx_cache_get(struct mlx5e_rq *rq, union mlx5e_alloc_uni
 
 	addr = page_pool_get_dma_addr(au->page);
 	/* Non-XSK always uses PAGE_SIZE. */
-	dma_sync_single_for_device(rq->pdev, addr, PAGE_SIZE, DMA_FROM_DEVICE);
+	dma_sync_single_for_device(rq->pdev, addr, PAGE_SIZE, rq->buff.map_dir);
 	return true;
 }
 
@@ -282,8 +282,7 @@ static inline int mlx5e_page_alloc_pool(struct mlx5e_rq *rq, union mlx5e_alloc_u
 		return -ENOMEM;
 
 	/* Non-XSK always uses PAGE_SIZE. */
-	addr = dma_map_page_attrs(rq->pdev, au->page, 0, PAGE_SIZE,
-				  rq->buff.map_dir, DMA_ATTR_SKIP_CPU_SYNC);
+	addr = dma_map_page(rq->pdev, au->page, 0, PAGE_SIZE, rq->buff.map_dir);
 	if (unlikely(dma_mapping_error(rq->pdev, addr))) {
 		page_pool_recycle_direct(rq->page_pool, au->page);
 		au->page = NULL;
@@ -427,14 +426,15 @@ mlx5e_add_skb_frag(struct mlx5e_rq *rq, struct sk_buff *skb,
 {
 	dma_addr_t addr = page_pool_get_dma_addr(au->page);
 
-	dma_sync_single_for_cpu(rq->pdev, addr + frag_offset, len, DMA_FROM_DEVICE);
+	dma_sync_single_for_cpu(rq->pdev, addr + frag_offset, len,
+				rq->buff.map_dir);
 	page_ref_inc(au->page);
 	skb_add_rx_frag(skb, skb_shinfo(skb)->nr_frags,
 			au->page, frag_offset, len, truesize);
 }
 
 static inline void
-mlx5e_copy_skb_header(struct device *pdev, struct sk_buff *skb,
+mlx5e_copy_skb_header(struct mlx5e_rq *rq, struct sk_buff *skb,
 		      struct page *page, dma_addr_t addr,
 		      int offset_from, int dma_offset, u32 headlen)
 {
@@ -442,7 +442,8 @@ mlx5e_copy_skb_header(struct device *pdev, struct sk_buff *skb,
 	/* Aligning len to sizeof(long) optimizes memcpy performance */
 	unsigned int len = ALIGN(headlen, sizeof(long));
 
-	dma_sync_single_for_cpu(pdev, addr + dma_offset, len, DMA_FROM_DEVICE);
+	dma_sync_single_for_cpu(rq->pdev, addr + dma_offset, len,
+				rq->buff.map_dir);
 	skb_copy_to_linear_data(skb, from, len);
 }
 
@@ -1538,7 +1539,7 @@ mlx5e_skb_from_cqe_linear(struct mlx5e_rq *rq, struct mlx5e_wqe_frag_info *wi,
 
 	addr = page_pool_get_dma_addr(au->page);
 	dma_sync_single_range_for_cpu(rq->pdev, addr, wi->offset,
-				      frag_size, DMA_FROM_DEVICE);
+				      frag_size, rq->buff.map_dir);
 	net_prefetch(data);
 
 	prog = rcu_dereference(rq->xdp_prog);
@@ -1587,7 +1588,7 @@ mlx5e_skb_from_cqe_nonlinear(struct mlx5e_rq *rq, struct mlx5e_wqe_frag_info *wi
 
 	addr = page_pool_get_dma_addr(au->page);
 	dma_sync_single_range_for_cpu(rq->pdev, addr, wi->offset,
-				      rq->buff.frame0_sz, DMA_FROM_DEVICE);
+				      rq->buff.frame0_sz, rq->buff.map_dir);
 	net_prefetchw(va); /* xdp_frame data area */
 	net_prefetch(va + rx_headroom);
 
@@ -1608,7 +1609,7 @@ mlx5e_skb_from_cqe_nonlinear(struct mlx5e_rq *rq, struct mlx5e_wqe_frag_info *wi
 
 		addr = page_pool_get_dma_addr(au->page);
 		dma_sync_single_for_cpu(rq->pdev, addr + wi->offset,
-					frag_consumed_bytes, DMA_FROM_DEVICE);
+					frag_consumed_bytes, rq->buff.map_dir);
 
 		if (!xdp_buff_has_frags(&xdp)) {
 			/* Init on the first fragment to avoid cold cache access
@@ -1905,7 +1906,7 @@ mlx5e_skb_from_cqe_mpwrq_nonlinear(struct mlx5e_rq *rq, struct mlx5e_mpw_info *w
 	mlx5e_fill_skb_data(skb, rq, au, byte_cnt, frag_offset);
 	/* copy header */
 	addr = page_pool_get_dma_addr(head_au->page);
-	mlx5e_copy_skb_header(rq->pdev, skb, head_au->page, addr,
+	mlx5e_copy_skb_header(rq, skb, head_au->page, addr,
 			      head_offset, head_offset, headlen);
 	/* skb linear part was allocated with headlen and aligned to long */
 	skb->tail += headlen;
@@ -1939,7 +1940,7 @@ mlx5e_skb_from_cqe_mpwrq_linear(struct mlx5e_rq *rq, struct mlx5e_mpw_info *wi,
 
 	addr = page_pool_get_dma_addr(au->page);
 	dma_sync_single_range_for_cpu(rq->pdev, addr, head_offset,
-				      frag_size, DMA_FROM_DEVICE);
+				      frag_size, rq->buff.map_dir);
 	net_prefetch(data);
 
 	prog = rcu_dereference(rq->xdp_prog);
@@ -1987,7 +1988,7 @@ mlx5e_skb_from_cqe_shampo(struct mlx5e_rq *rq, struct mlx5e_mpw_info *wi,
 
 	if (likely(frag_size <= BIT(MLX5E_SHAMPO_LOG_MAX_HEADER_ENTRY_SIZE))) {
 		/* build SKB around header */
-		dma_sync_single_range_for_cpu(rq->pdev, head->addr, 0, frag_size, DMA_FROM_DEVICE);
+		dma_sync_single_range_for_cpu(rq->pdev, head->addr, 0, frag_size, rq->buff.map_dir);
 		prefetchw(hdr);
 		prefetch(data);
 		skb = mlx5e_build_linear_skb(rq, hdr, frag_size, rx_headroom, head_size, 0);
@@ -2009,7 +2010,7 @@ mlx5e_skb_from_cqe_shampo(struct mlx5e_rq *rq, struct mlx5e_mpw_info *wi,
 		}
 
 		prefetchw(skb->data);
-		mlx5e_copy_skb_header(rq->pdev, skb, head->page, head->addr,
+		mlx5e_copy_skb_header(rq, skb, head->page, head->addr,
 				      head_offset + rx_headroom,
 				      rx_headroom, head_size);
 		/* skb linear part was allocated with headlen and aligned to long */
Signature · 1 line
2.38.1
#12 lore raw

PATCH v2 9/11 net/mlx5e: TC, Fix wrong rejection of packet-per-second policing

In reply to Saeed Mahameed

From: Jianbo Liu <[email protected]>

In the bellow commit, we added support for PPS policing without removing the check which block offload of such cases. Fix it by removing this check.

Fixes a8d52b024d6d ("net/mlx5e: TC, Support offloading police action")
---
 drivers/net/ethernet/mellanox/mlx5/core/en_tc.c | 6 ------
 1 file changed, 6 deletions(-)
drivers/net/ethernet/mellanox/mlx5/core/en_tc.c +0 -6
index dd6fea9e9a5b..372dfb89e396 100644
@@ -4758,12 +4758,6 @@ int mlx5e_policer_validate(const struct flow_action *action,
 		return -EOPNOTSUPP;
 	}
 
-	if (act->police.rate_pkt_ps) {
-		NL_SET_ERR_MSG_MOD(extack,
-				   "QoS offload not support packets per second");
-		return -EOPNOTSUPP;
-	}
-
 	return 0;
 }
 
Signature · 1 line
2.38.1
#11 lore raw

PATCH v2 8/11 net/mlx5e: Fix tc acts array not to be dependent on enum order

In reply to Saeed Mahameed

From: Roi Dayan <[email protected]>

The tc acts array should not be dependent on kernel internal flow action id enum. Fix the array initialization.

Fixes fad547906980 ("net/mlx5e: Add tc action infrastructure")
---
 .../mellanox/mlx5/core/en/tc/act/act.c        | 92 +++++++------------
 1 file changed, 32 insertions(+), 60 deletions(-)
drivers/net/ethernet/mellanox/mlx5/core/en/tc/act/act.c +32 -60
index 305fde62a78d..3337241cfd84 100644
@@ -6,70 +6,42 @@
 #include "en/tc_priv.h"
 #include "mlx5_core.h"
 
-/* Must be aligned with enum flow_action_id. */
 static struct mlx5e_tc_act *tc_acts_fdb[NUM_FLOW_ACTIONS] = {
-	&mlx5e_tc_act_accept,
-	&mlx5e_tc_act_drop,
-	&mlx5e_tc_act_trap,
-	&mlx5e_tc_act_goto,
-	&mlx5e_tc_act_mirred,
-	&mlx5e_tc_act_mirred,
-	&mlx5e_tc_act_redirect_ingress,
-	NULL, /* FLOW_ACTION_MIRRED_INGRESS, */
-	&mlx5e_tc_act_vlan,
-	&mlx5e_tc_act_vlan,
-	&mlx5e_tc_act_vlan_mangle,
-	&mlx5e_tc_act_tun_encap,
-	&mlx5e_tc_act_tun_decap,
-	&mlx5e_tc_act_pedit,
-	&mlx5e_tc_act_pedit,
-	&mlx5e_tc_act_csum,
-	NULL, /* FLOW_ACTION_MARK, */
-	&mlx5e_tc_act_ptype,
-	NULL, /* FLOW_ACTION_PRIORITY, */
-	NULL, /* FLOW_ACTION_WAKE, */
-	NULL, /* FLOW_ACTION_QUEUE, */
-	&mlx5e_tc_act_sample,
-	&mlx5e_tc_act_police,
-	&mlx5e_tc_act_ct,
-	NULL, /* FLOW_ACTION_CT_METADATA, */
-	&mlx5e_tc_act_mpls_push,
-	&mlx5e_tc_act_mpls_pop,
-	NULL, /* FLOW_ACTION_MPLS_MANGLE, */
-	NULL, /* FLOW_ACTION_GATE, */
-	NULL, /* FLOW_ACTION_PPPOE_PUSH, */
-	NULL, /* FLOW_ACTION_JUMP, */
-	NULL, /* FLOW_ACTION_PIPE, */
-	&mlx5e_tc_act_vlan,
-	&mlx5e_tc_act_vlan,
+	[FLOW_ACTION_ACCEPT] = &mlx5e_tc_act_accept,
+	[FLOW_ACTION_DROP] = &mlx5e_tc_act_drop,
+	[FLOW_ACTION_TRAP] = &mlx5e_tc_act_trap,
+	[FLOW_ACTION_GOTO] = &mlx5e_tc_act_goto,
+	[FLOW_ACTION_REDIRECT] = &mlx5e_tc_act_mirred,
+	[FLOW_ACTION_MIRRED] = &mlx5e_tc_act_mirred,
+	[FLOW_ACTION_REDIRECT_INGRESS] = &mlx5e_tc_act_redirect_ingress,
+	[FLOW_ACTION_VLAN_PUSH] = &mlx5e_tc_act_vlan,
+	[FLOW_ACTION_VLAN_POP] = &mlx5e_tc_act_vlan,
+	[FLOW_ACTION_VLAN_MANGLE] = &mlx5e_tc_act_vlan_mangle,
+	[FLOW_ACTION_TUNNEL_ENCAP] = &mlx5e_tc_act_tun_encap,
+	[FLOW_ACTION_TUNNEL_DECAP] = &mlx5e_tc_act_tun_decap,
+	[FLOW_ACTION_MANGLE] = &mlx5e_tc_act_pedit,
+	[FLOW_ACTION_ADD] = &mlx5e_tc_act_pedit,
+	[FLOW_ACTION_CSUM] = &mlx5e_tc_act_csum,
+	[FLOW_ACTION_PTYPE] = &mlx5e_tc_act_ptype,
+	[FLOW_ACTION_SAMPLE] = &mlx5e_tc_act_sample,
+	[FLOW_ACTION_POLICE] = &mlx5e_tc_act_police,
+	[FLOW_ACTION_CT] = &mlx5e_tc_act_ct,
+	[FLOW_ACTION_MPLS_PUSH] = &mlx5e_tc_act_mpls_push,
+	[FLOW_ACTION_MPLS_POP] = &mlx5e_tc_act_mpls_pop,
+	[FLOW_ACTION_VLAN_PUSH_ETH] = &mlx5e_tc_act_vlan,
+	[FLOW_ACTION_VLAN_POP_ETH] = &mlx5e_tc_act_vlan,
 };
 
-/* Must be aligned with enum flow_action_id. */
 static struct mlx5e_tc_act *tc_acts_nic[NUM_FLOW_ACTIONS] = {
-	&mlx5e_tc_act_accept,
-	&mlx5e_tc_act_drop,
-	NULL, /* FLOW_ACTION_TRAP, */
-	&mlx5e_tc_act_goto,
-	&mlx5e_tc_act_mirred_nic,
-	NULL, /* FLOW_ACTION_MIRRED, */
-	NULL, /* FLOW_ACTION_REDIRECT_INGRESS, */
-	NULL, /* FLOW_ACTION_MIRRED_INGRESS, */
-	NULL, /* FLOW_ACTION_VLAN_PUSH, */
-	NULL, /* FLOW_ACTION_VLAN_POP, */
-	NULL, /* FLOW_ACTION_VLAN_MANGLE, */
-	NULL, /* FLOW_ACTION_TUNNEL_ENCAP, */
-	NULL, /* FLOW_ACTION_TUNNEL_DECAP, */
-	&mlx5e_tc_act_pedit,
-	&mlx5e_tc_act_pedit,
-	&mlx5e_tc_act_csum,
-	&mlx5e_tc_act_mark,
-	NULL, /* FLOW_ACTION_PTYPE, */
-	NULL, /* FLOW_ACTION_PRIORITY, */
-	NULL, /* FLOW_ACTION_WAKE, */
-	NULL, /* FLOW_ACTION_QUEUE, */
-	NULL, /* FLOW_ACTION_SAMPLE, */
-	NULL, /* FLOW_ACTION_POLICE, */
-	&mlx5e_tc_act_ct,
+	[FLOW_ACTION_ACCEPT] = &mlx5e_tc_act_accept,
+	[FLOW_ACTION_DROP] = &mlx5e_tc_act_drop,
+	[FLOW_ACTION_GOTO] = &mlx5e_tc_act_goto,
+	[FLOW_ACTION_REDIRECT] = &mlx5e_tc_act_mirred_nic,
+	[FLOW_ACTION_MANGLE] = &mlx5e_tc_act_pedit,
+	[FLOW_ACTION_ADD] = &mlx5e_tc_act_pedit,
+	[FLOW_ACTION_CSUM] = &mlx5e_tc_act_csum,
+	[FLOW_ACTION_MARK] = &mlx5e_tc_act_mark,
+	[FLOW_ACTION_CT] = &mlx5e_tc_act_ct,
 };
 
 /**
Signature · 1 line
2.38.1
#13 lore raw

PATCH v2 10/11 net/mlx5e: E-Switch, Fix comparing termination table instance

In reply to Saeed Mahameed

From: Roi Dayan <[email protected]>

The pkt_reformat pointer being saved under flow_act and not dest attribute in the termination table instance. Fix the comparison pointers.

Also fix returning success if one pkt_reformat pointer is null and the other is not.

Fixes 249ccc3c95bd ("net/mlx5e: Add support for offloading traffic from uplink to uplink")
---
 .../mellanox/mlx5/core/eswitch_offloads_termtbl.c  | 14 ++++++++------
 1 file changed, 8 insertions(+), 6 deletions(-)
drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads_termtbl.c +8 -6
index ee568bf34ae2..108a3503f413 100644
@@ -30,9 +30,9 @@ mlx5_eswitch_termtbl_hash(struct mlx5_flow_act *flow_act,
 		     sizeof(dest->vport.num), hash);
 	hash = jhash((const void *)&dest->vport.vhca_id,
 		     sizeof(dest->vport.num), hash);
-	if (dest->vport.pkt_reformat)
-		hash = jhash(dest->vport.pkt_reformat,
-			     sizeof(*dest->vport.pkt_reformat),
+	if (flow_act->pkt_reformat)
+		hash = jhash(flow_act->pkt_reformat,
+			     sizeof(*flow_act->pkt_reformat),
 			     hash);
 	return hash;
 }
@@ -53,9 +53,11 @@ mlx5_eswitch_termtbl_cmp(struct mlx5_flow_act *flow_act1,
 	if (ret)
 		return ret;
 
-	return dest1->vport.pkt_reformat && dest2->vport.pkt_reformat ?
-	       memcmp(dest1->vport.pkt_reformat, dest2->vport.pkt_reformat,
-		      sizeof(*dest1->vport.pkt_reformat)) : 0;
+	if (flow_act1->pkt_reformat && flow_act2->pkt_reformat)
+		return memcmp(flow_act1->pkt_reformat, flow_act2->pkt_reformat,
+			      sizeof(*flow_act1->pkt_reformat));
+
+	return !(flow_act1->pkt_reformat == flow_act2->pkt_reformat);
 }
 
 static int
Signature · 1 line
2.38.1
#14 lore raw

PATCH v2 11/11 net/mlx5e: TC, Fix slab-out-of-bounds in parse_tc_actions

In reply to Saeed Mahameed

From: Roi Dayan <[email protected]>

esw_attr is only allocated if namespace is fdb.

BUG: KASAN: slab-out-of-bounds in parse_tc_actions+0xdc6/0x10e0 [mlx5_core]
Write of size 4 at addr ffff88815f185b04 by task tc/2135

CPU: 5 PID: 2135 Comm: tc Not tainted 6.1.0-rc2+ #2
Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.13.0-0-gf21b5a4aeb02-prebuilt.qemu.org 04/01/2014
Call Trace:
 <TASK>
 dump_stack_lvl+0x57/0x7d
 print_report+0x170/0x471
 ? parse_tc_actions+0xdc6/0x10e0 [mlx5_core]
 kasan_report+0xbc/0xf0
 ? parse_tc_actions+0xdc6/0x10e0 [mlx5_core]
 parse_tc_actions+0xdc6/0x10e0 [mlx5_core]
Fixes 94d651739e17 ("net/mlx5e: TC, Fix cloned flow attr instance dests are not zeroed")
---
 drivers/net/ethernet/mellanox/mlx5/core/en_tc.c | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)
drivers/net/ethernet/mellanox/mlx5/core/en_tc.c +6 -2
index 372dfb89e396..5a6aa61ec82a 100644
@@ -3633,10 +3633,14 @@ mlx5e_clone_flow_attr_for_post_act(struct mlx5_flow_attr *attr,
 	attr2->action = 0;
 	attr2->flags = 0;
 	attr2->parse_attr = parse_attr;
-	attr2->esw_attr->out_count = 0;
-	attr2->esw_attr->split_count = 0;
 	attr2->dest_chain = 0;
 	attr2->dest_ft = NULL;
+
+	if (ns_type == MLX5_FLOW_NAMESPACE_FDB) {
+		attr2->esw_attr->out_count = 0;
+		attr2->esw_attr->split_count = 0;
+	}
+
 	return attr2;
 }
 
Signature · 1 line
2.38.1
maintainer
#6 lore raw

Re: [V2 net 05/11] net/mlx5: Fix possible deadlock on mlx5e_tx_timeout_work

In reply to Saeed Mahameed

drivers/net/ethernet/mellanox/mlx5/core/en_main.c static int mlx5e_switch_priv_channels(struct mlx5e_priv *priv, · line 2959

Quoted 3 lines · On Sat, 5 Nov 2022 00:10:22 -0700 Saeed Mahameed wrote:
drivers/net/ethernet/mellanox/mlx5/core/en_main.c +3 -0
+	/* Once deactivated, new tx_timeout_work won't be initiated. */
+	if (current_work() != &priv->tx_timeout_work)
+		cancel_work_sync(&priv->tx_timeout_work);

The work takes rtnl_lock, are there no callers of mlx5e_switch_priv_channels() that are under rtnl_lock()?

This patch is definitely going onto my "expecting Fixes" bingo card :S

maintainer
#7 lore raw

Re: [V2 net 05/11] net/mlx5: Fix possible deadlock on mlx5e_tx_timeout_work

Quoted 10 lines · Jakub Kicinski

drivers/net/ethernet/mellanox/mlx5/core/en_main.c static int mlx5e_switch_priv_channels(struct mlx5e_priv *priv, · line 2959

Quoted 3 lines · On Sat, 5 Nov 2022 00:10:22 -0700 Saeed Mahameed wrote:
drivers/net/ethernet/mellanox/mlx5/core/en_main.c +3 -0
+	/* Once deactivated, new tx_timeout_work won't be initiated. */
+	if (current_work() != &priv->tx_timeout_work)
+		cancel_work_sync(&priv->tx_timeout_work);

The work takes rtnl_lock, are there no callers of mlx5e_switch_priv_channels() that are under rtnl_lock()?

This patch is definitely going onto my "expecting Fixes" bingo card :S

I think Jakub is right and even mlx5e_close_locked() will deadlock on cancel_work_sync() if the work is scheduled but it has not yet acquired the rtnl lock.

IIRC lockdep is not able to catch this kind of situation, so you can only observe the deadlock when reaching the critical scenario.

I'm wild guessing than a possible solution would be restrict the state_lock scope in mlx5e_tx_timeout_work() around the state check, without additional cancel_work operations.

Thanks,

Paolo

#8 lore raw

Re: [V2 net 05/11] net/mlx5: Fix possible deadlock on mlx5e_tx_timeout_work

Quoted 9 lines · Paolo Abeni
Quoted 8 lines · Jakub Kicinski

drivers/net/ethernet/mellanox/mlx5/core/en_main.c static int mlx5e_switch_priv_channels(struct mlx5e_priv *priv, · line 2959

Quoted 3 lines · On Sat, 5 Nov 2022 00:10:22 -0700 Saeed Mahameed wrote:
drivers/net/ethernet/mellanox/mlx5/core/en_main.c +3 -0
+	/* Once deactivated, new tx_timeout_work won't be initiated. */
+	if (current_work() != &priv->tx_timeout_work)
+		cancel_work_sync(&priv->tx_timeout_work);

The work takes rtnl_lock, are there no callers of mlx5e_switch_priv_channels() that are under rtnl_lock()?

Quoted 6 lines
Quoted 2 lines

This patch is definitely going onto my "expecting Fixes" bingo card :S

I think Jakub is right and even mlx5e_close_locked() will deadlock on cancel_work_sync() if the work is scheduled but it has not yet acquired the rtnl lock.

Yes you are absolutely correct, you can see the deadlock just by looking at the patch diff and applying common sense that mlx5e_switch_priv_channels() is being called under rtnl.

Quoted 8 lines

IIRC lockdep is not able to catch this kind of situation, so you can only observe the deadlock when reaching the critical scenario.

I'm wild guessing than a possible solution would be restrict the state_lock scope in mlx5e_tx_timeout_work() around the state check, without additional cancel_work operations.

Thanks, i will drop the patch for now and send v3 without it.

Quoted 4 lines

Thanks,

Paolo