Take devlink lock on mlx4 and mlx5 callbacks
v2 cover Take devlink lock on mlx4 and mlx5 callbacks
Prepare mlx4 and mlx5 drivers to have all devlink callbacks called with devlink instance locked. Change mlx4 driver to use devl_ API where needed to have devlink reload callbacks locked. Change mlx5 driver to use devl_ API where needed to have devlink reload and devlink health callbacks locked.
As mlx5 is the only driver which needed changes to enable calling health callbacks with devlink instance locked, this patchset also removes DEVLINK_NL_FLAG_NO_LOCK flag from devlink health callbacks.
This patchset will be followed by a patchset that will remove DEVLINK_NL_FLAG_NO_LOCK flag from devlink and will remove devlink_mutex.
Jiri Pirko (2):
net: devlink: remove region snapshot ID tracking dependency on
devlink->lock
net: devlink: remove region snapshots list dependency on devlink->lock
Moshe Shemesh (7):
net/mlx5: Move fw reset unload to mlx5_fw_reset_complete_reload net/mlx5: Lock mlx5 devlink reload callbacks net/mlx4: Use devl_ API for devlink region create / destroy net/mlx4: Use devl_ API for devlink port register / unregister net/mlx4: Lock mlx4 devlink reload callback
net/mlx5: Lock mlx5 devlink health recovery callback
devlink: Hold the instance lock in health callbacks
drivers/net/ethernet/mellanox/mlx4/catas.c | 5 + drivers/net/ethernet/mellanox/mlx4/crdump.c | 20 +-- drivers/net/ethernet/mellanox/mlx4/main.c | 44 +++++- drivers/net/ethernet/mellanox/mlx5/core/dev.c | 19 +-- .../net/ethernet/mellanox/mlx5/core/devlink.c | 59 +++++--- .../net/ethernet/mellanox/mlx5/core/eswitch.c | 18 +-- .../ethernet/mellanox/mlx5/core/fw_reset.c | 10 +- .../net/ethernet/mellanox/mlx5/core/health.c | 4 + .../net/ethernet/mellanox/mlx5/core/main.c | 38 ++++- .../ethernet/mellanox/mlx5/core/mlx5_core.h | 2 + .../net/ethernet/mellanox/mlx5/core/sriov.c | 6 + net/core/devlink.c | 135 ++++++++++-------- 12 files changed, 227 insertions(+), 133 deletions(-)
Signature · 1 line
2.18.2
PATCH v2 1/9 net: devlink: remove region snapshot ID tracking dependency on devlink->lock
From: Jiri Pirko <[email protected]>
After mlx4 driver is converted to do locked reload, functions to get/put regions snapshot ID may be called from both locked and unlocked context.
So resolve this by removing dependency on devlink->lock for region snapshot ID tracking by using internal xa_lock() to maintain shapshot_ids xa_array consistency.
---v1->v2:
- changed GFP_KERNEL to GFP_ATOMIC in __xa_store() calls
--- net/core/devlink.c | 64 ++++++++++++++++++++++++---------------------- 1 file changed, 33 insertions(+), 31 deletions(-)
net/core/devlink.c +33 -31
@@ -5690,21 +5690,28 @@ static int __devlink_snapshot_id_increment(struct devlink *devlink, u32 id) {unsigned long count;void *p; +int err; -devl_assert_locked (devlink); - +xa_lock (&devlink->snapshot_ids); p =xa_load (&devlink->snapshot_ids, id); -if (WARN_ON (!p)) -return -EINVAL ; +if (WARN_ON (!p)) { + err = -EINVAL ; + goto unlock; + } -if (WARN_ON (!xa_is_value (p))) -return -EINVAL ; +if (WARN_ON (!xa_is_value (p))) { + err = -EINVAL ; + goto unlock; + } count =xa_to_value (p); count++; -return xa_err (xa_store (&devlink->snapshot_ids, id,xa_mk_value (count), -GFP_KERNEL )); + err =xa_err (__xa_store (&devlink->snapshot_ids, id,xa_mk_value (count), +GFP_ATOMIC )); +unlock: +xa_unlock (&devlink->snapshot_ids); +return err; }/**
@@ -5727,25 +5734,26 @@ static void __devlink_snapshot_id_decrement(struct devlink *devlink, u32 id)unsigned long count;void *p; -devl_assert_locked (devlink); - +xa_lock (&devlink->snapshot_ids); p =xa_load (&devlink->snapshot_ids, id);if (WARN_ON (!p)) -return ; + goto unlock;if (WARN_ON (!xa_is_value (p))) -return ; + goto unlock; count =xa_to_value (p);if (count >1 ) { count--; -xa_store (&devlink->snapshot_ids, id,xa_mk_value (count), -GFP_KERNEL ); +__xa_store (&devlink->snapshot_ids, id,xa_mk_value (count), +GFP_ATOMIC ); }else {/* If this was the last user, we can erase this id */ -xa_erase (&devlink->snapshot_ids, id); +__xa_erase (&devlink->snapshot_ids, id); } +unlock: +xa_unlock (&devlink->snapshot_ids); }/**
@@ -5766,13 +5774,17 @@ static void __devlink_snapshot_id_decrement(struct devlink *devlink, u32 id)*/ static int __devlink_snapshot_id_insert (struct devlink *devlink,u32 id) { -devl_assert_locked (devlink); +int err; -if (xa_load (&devlink->snapshot_ids, id)) +xa_lock (&devlink->snapshot_ids); +if (xa_load (&devlink->snapshot_ids, id)) { +xa_unlock (&devlink->snapshot_ids);return -EEXIST ; - -return xa_err (xa_store (&devlink->snapshot_ids, id,xa_mk_value (0 ), -GFP_KERNEL )); + } + err =xa_err (__xa_store (&devlink->snapshot_ids, id,xa_mk_value (0 ), +GFP_ATOMIC )); +xa_unlock (&devlink->snapshot_ids); +return err; }/**
@@ -5793,8 +5805,6 @@ static int __devlink_snapshot_id_insert(struct devlink *devlink, u32 id)*/ static int __devlink_region_snapshot_id_get (struct devlink *devlink,u32 *id) { -devl_assert_locked (devlink); -return xa_alloc (&devlink->snapshot_ids, id,xa_mk_value (1 ), xa_limit_32b,GFP_KERNEL ); }
@@ -11226,13 +11236,7 @@ EXPORT_SYMBOL_GPL(devlink_region_destroy);*/ int devlink_region_snapshot_id_get (struct devlink *devlink,u32 *id) { -int err; - -devl_lock (devlink); - err =__devlink_region_snapshot_id_get (devlink, id); -devl_unlock (devlink); - -return err; +return __devlink_region_snapshot_id_get (devlink, id); }EXPORT_SYMBOL_GPL (devlink_region_snapshot_id_get);
@@ -11248,9 +11252,7 @@ EXPORT_SYMBOL_GPL(devlink_region_snapshot_id_get);*/ void devlink_region_snapshot_id_put (struct devlink *devlink,u32 id) { -devl_lock (devlink);__devlink_snapshot_id_decrement (devlink, id); -devl_unlock (devlink); }EXPORT_SYMBOL_GPL (devlink_region_snapshot_id_put);
Signature · 1 line
2.18.2
PATCH v2 2/9 net: devlink: remove region snapshots list dependency on devlink->lock
In reply to Moshe ShemeshFrom: Jiri Pirko <[email protected]>
After mlx4 driver is converted to do locked reload, devlink_region_snapshot_create() may be called from both locked and unlocked context.
Note that in mlx4 region snapshots could be created on any command failure. That can happen in any flow that involves commands to FW, which means most of the driver flows.
So resolve this by removing dependency on devlink->lock for region snapshots list consistency and introduce new mutex to ensure it.
---v1->v2:
- enhanced patch description
--- net/core/devlink.c | 41 +++++++++++++++++++++++++++++------------ 1 file changed, 29 insertions(+), 12 deletions(-)
net/core/devlink.c +29 -12
@@ -695,6 +695,10 @@ struct devlink_region {const struct devlink_region_ops *ops;const struct devlink_port_region_ops *port_ops; }; +struct mutex snapshot_lock;/* protects snapshot_list, +* max_snapshots and cur_snapshots +* consistency. +*/ struct list_head snapshot_list;u32 max_snapshots;u32 cur_snapshots;
@@ -5817,7 +5821,7 @@ static int __devlink_region_snapshot_id_get(struct devlink *devlink, u32 *id)* Multiple snapshots can be created on a region. * The @snapshot_id should be obtained using the getter function. * -* Must be called only while holding the devlink instance lock. +* Must be called only while holding the region snapshot lock. * * @region: devlink region of the snapshot * @data: snapshot data
@@ -5831,7 +5835,7 @@ __devlink_region_snapshot_create(struct devlink_region *region,struct devlink_snapshot *snapshot;int err; -devl_assert_locked (devlink); +lockdep_assert_held (®ion->snapshot_lock);/* check if region can hold one more snapshot */ if (region->cur_snapshots == region->max_snapshots)
@@ -5869,7 +5873,7 @@ static void devlink_region_snapshot_del(struct devlink_region *region, {struct devlink *devlink = region->devlink; -devl_assert_locked (devlink); +lockdep_assert_held (®ion->snapshot_lock);devlink_nl_region_notify (region, snapshot,DEVLINK_CMD_REGION_DEL ); region->cur_snapshots--;
@@ -6048,11 +6052,15 @@ static int devlink_nl_cmd_region_del(struct sk_buff *skb,if (!region)return -EINVAL ; +mutex_lock (®ion->snapshot_lock); snapshot =devlink_region_snapshot_get_by_id (region, snapshot_id); -if (!snapshot) +if (!snapshot) { +mutex_unlock (®ion->snapshot_lock);return -EINVAL ; + }devlink_region_snapshot_del (region, snapshot); +mutex_unlock (®ion->snapshot_lock);return 0 ; }
@@ -6100,9 +6108,12 @@ devlink_nl_cmd_region_new(struct sk_buff *skb, struct genl_info *info)return -EOPNOTSUPP ; } +mutex_lock (®ion->snapshot_lock); +if (region->cur_snapshots == region->max_snapshots) {NL_SET_ERR_MSG_MOD (info->extack,"The region has reached the maximum number of stored snapshots" ); -return -ENOSPC ; + err = -ENOSPC ; + goto unlock; } snapshot_id_attr = info->attrs[DEVLINK_ATTR_REGION_SNAPSHOT_ID ];
@@ -6111,17 +6122,18 @@ devlink_nl_cmd_region_new(struct sk_buff *skb, struct genl_info *info)if (devlink_region_snapshot_get_by_id (region, snapshot_id)) {NL_SET_ERR_MSG_MOD (info->extack,"The requested snapshot id is already in use" ); -return -EEXIST ; + err = -EEXIST ; + goto unlock; } err =__devlink_snapshot_id_insert (devlink, snapshot_id);if (err) -return err; + goto unlock; }else { err =__devlink_region_snapshot_id_get (devlink, &snapshot_id);if (err) {NL_SET_ERR_MSG_MOD (info->extack,"Failed to allocate a new snapshot id" ); -return err; + goto unlock; } }
@@ -6159,16 +6171,20 @@ devlink_nl_cmd_region_new(struct sk_buff *skb, struct genl_info *info) goto err_notify; } +mutex_unlock (®ion->snapshot_lock);return 0 ; err_snapshot_create: region->ops->destructor (data); err_snapshot_capture:__devlink_snapshot_id_decrement (devlink, snapshot_id); +mutex_unlock (®ion->snapshot_lock);return err; err_notify:devlink_region_snapshot_del (region, snapshot); +unlock: +mutex_unlock (®ion->snapshot_lock);return err; }
@@ -11094,6 +11110,7 @@ struct devlink_region *devl_region_create(struct devlink *devlink, region->ops = ops; region->size = region_size;INIT_LIST_HEAD (®ion->snapshot_list); +mutex_init (®ion->snapshot_lock);list_add_tail (®ion->list, &devlink->region_list);devlink_nl_region_notify (region,NULL ,DEVLINK_CMD_REGION_NEW );
@@ -11167,6 +11184,7 @@ devlink_port_region_create(struct devlink_port *port, region->port_ops = ops; region->size = region_size;INIT_LIST_HEAD (®ion->snapshot_list); +mutex_init (®ion->snapshot_lock);list_add_tail (®ion->list, &port->region_list);devlink_nl_region_notify (region,NULL ,DEVLINK_CMD_REGION_NEW );
@@ -11196,6 +11214,7 @@ void devl_region_destroy(struct devlink_region *region)devlink_region_snapshot_del (region, snapshot);list_del (®ion->list); +mutex_destroy (®ion->snapshot_lock);devlink_nl_region_notify (region,NULL ,DEVLINK_CMD_REGION_DEL );kfree (region);
@@ -11271,13 +11290,11 @@ EXPORT_SYMBOL_GPL(devlink_region_snapshot_id_put);int devlink_region_snapshot_create (struct devlink_region *region,u8 *data,u32 snapshot_id) { -struct devlink *devlink = region->devlink;int err; -devl_lock (devlink); +mutex_lock (®ion->snapshot_lock); err =__devlink_region_snapshot_create (region, data, snapshot_id); -devl_unlock (devlink); - +mutex_unlock (®ion->snapshot_lock);return err; }EXPORT_SYMBOL_GPL (devlink_region_snapshot_create);
Signature · 1 line
2.18.2
PATCH v2 3/9 net/mlx5: Move fw reset unload to mlx5_fw_reset_complete_reload
In reply to Moshe ShemeshRefactor fw reset code to have the unload driver part done on mlx5_fw_reset_complete_reload(), so if it was called by the PF which initiated the reload fw activate flow, the unload part will be handled by the mlx5_devlink_reload_fw_activate() callback itself and not by the reset event work.
This will be used by the downstream patch to invoke devlink reload callbacks with devlink lock held.
--- drivers/net/ethernet/mellanox/mlx5/core/devlink.c | 11 ++++++++++- drivers/net/ethernet/mellanox/mlx5/core/fw_reset.c | 10 +++------- 2 files changed, 13 insertions(+), 8 deletions(-)
drivers/net/ethernet/mellanox/mlx5/core/devlink.c +10 -1
@@ -104,7 +104,16 @@ static int mlx5_devlink_reload_fw_activate(struct devlink *devlink, struct netliif (err)return err; -return mlx5_fw_reset_wait_reset_done (dev); + err =mlx5_fw_reset_wait_reset_done (dev); +if (err) +return err; + +mlx5_unload_one (dev); + err =mlx5_health_wait_pci_up (dev); +if (err) +NL_SET_ERR_MSG_MOD (extack,"FW activate aborted, PCI reads fail after reset" ); + +return err; } static int mlx5_devlink_trigger_fw_live_patch(struct devlink *devlink,
drivers/net/ethernet/mellanox/mlx5/core/fw_reset.c +3 -7
@@ -149,6 +149,9 @@ static void mlx5_fw_reset_complete_reload(struct mlx5_core_dev *dev)if (test_bit (MLX5_FW_RESET_FLAGS_PENDING_COMP , &fw_reset->reset_flags)) {complete (&fw_reset->done); }else { +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); devlink_remote_reload_actions_performed(priv_to_devlink(dev), 0, BIT(DEVLINK_RELOAD_ACTION_DRIVER_REINIT) |
@@ -183,15 +186,9 @@ static void mlx5_sync_reset_reload_work(struct work_struct *work) struct mlx5_fw_reset *fw_reset = container_of(work, struct mlx5_fw_reset, reset_reload_work);struct mlx5_core_dev *dev = fw_reset->dev; -int err;mlx5_sync_reset_clear_reset_requested (dev, false);mlx5_enter_error_state (dev, true); -mlx5_unload_one (dev); - err =mlx5_health_wait_pci_up (dev); -if (err) -mlx5_core_err (dev,"reset reload flow aborted, PCI reads still not working\n" ); - fw_reset->ret = err;mlx5_fw_reset_complete_reload (dev); }
@@ -395,7 +392,6 @@ static void mlx5_sync_reset_now_event(struct work_struct *work) }mlx5_enter_error_state (dev, true); -mlx5_unload_one (dev); done: fw_reset->ret = err;mlx5_fw_reset_complete_reload (dev);
Signature · 1 line
2.18.2
PATCH v2 4/9 net/mlx5: Lock mlx5 devlink reload callbacks
In reply to Moshe ShemeshChange devlink instance locks in mlx5 driver to have devlink reload callbacks locked, while keeping all driver paths which lead to devl_ API functions called by the driver locked.
Add mlx5_load_one_devl_locked() and mlx5_unload_one_devl_locked() which are used by the paths which are already locked such as devlink reload callbacks.
This patch makes the driver use devl_ API also for traps register as these functions are called from the driver paths parallel to reload that requires locking now.
--- drivers/net/ethernet/mellanox/mlx5/core/dev.c | 19 ++----- .../net/ethernet/mellanox/mlx5/core/devlink.c | 50 ++++++++++++------- .../net/ethernet/mellanox/mlx5/core/eswitch.c | 18 ++----- .../net/ethernet/mellanox/mlx5/core/main.c | 34 ++++++++++++- .../ethernet/mellanox/mlx5/core/mlx5_core.h | 2 + .../net/ethernet/mellanox/mlx5/core/sriov.c | 6 +++ 6 files changed, 79 insertions(+), 50 deletions(-)
drivers/net/ethernet/mellanox/mlx5/core/dev.c +4 -15
@@ -335,13 +335,12 @@ static void del_adev(struct auxiliary_device *adev)int mlx5_attach_device (struct mlx5_core_dev *dev) { -struct devlink *devlink =priv_to_devlink (dev);struct mlx5_priv *priv = &dev->priv;struct auxiliary_device *adev;struct auxiliary_driver *adrv;int ret =0 , i; -devl_lock (devlink); +devl_assert_locked (priv_to_devlink (dev));mutex_lock (&mlx5_intf_mutex); priv->flags &= ~MLX5_PRIV_FLAGS_DETACH ; priv->flags |=MLX5_PRIV_FLAGS_MLX5E_LOCKED_FLOW ;
@@ -394,20 +393,18 @@ int mlx5_attach_device(struct mlx5_core_dev *dev) } priv->flags &= ~MLX5_PRIV_FLAGS_MLX5E_LOCKED_FLOW ;mutex_unlock (&mlx5_intf_mutex); -devl_unlock (devlink);return ret; }void mlx5_detach_device (struct mlx5_core_dev *dev) { -struct devlink *devlink =priv_to_devlink (dev);struct mlx5_priv *priv = &dev->priv;struct auxiliary_device *adev;struct auxiliary_driver *adrv;pm_message_t pm = {};int i; -devl_lock (devlink); +devl_assert_locked (priv_to_devlink (dev));mutex_lock (&mlx5_intf_mutex); priv->flags |=MLX5_PRIV_FLAGS_MLX5E_LOCKED_FLOW ;for (i =ARRAY_SIZE (mlx5_adev_devices) -1 ; i >=0 ; i--) {
@@ -441,21 +438,17 @@ void mlx5_detach_device(struct mlx5_core_dev *dev) priv->flags &= ~MLX5_PRIV_FLAGS_MLX5E_LOCKED_FLOW ; priv->flags |=MLX5_PRIV_FLAGS_DETACH ;mutex_unlock (&mlx5_intf_mutex); -devl_unlock (devlink); }int mlx5_register_device (struct mlx5_core_dev *dev) { -struct devlink *devlink;int ret; - devlink =priv_to_devlink (dev); -devl_lock (devlink); +devl_assert_locked (priv_to_devlink (dev));mutex_lock (&mlx5_intf_mutex); dev->priv.flags &= ~MLX5_PRIV_FLAGS_DISABLE_ALL_ADEV ; ret =mlx5_rescan_drivers_locked (dev);mutex_unlock (&mlx5_intf_mutex); -devl_unlock (devlink);if (ret)mlx5_unregister_device (dev);
@@ -464,15 +457,11 @@ int mlx5_register_device(struct mlx5_core_dev *dev)void mlx5_unregister_device (struct mlx5_core_dev *dev) { -struct devlink *devlink; - - devlink =priv_to_devlink (dev); -devl_lock (devlink); +devl_assert_locked (priv_to_devlink (dev));mutex_lock (&mlx5_intf_mutex); dev->priv.flags =MLX5_PRIV_FLAGS_DISABLE_ALL_ADEV ;mlx5_rescan_drivers_locked (dev);mutex_unlock (&mlx5_intf_mutex); -devl_unlock (devlink); }static int add_drivers (struct mlx5_core_dev *dev)
drivers/net/ethernet/mellanox/mlx5/core/devlink.c +31 -19
115 lines not on this page · raw message
drivers/net/ethernet/mellanox/mlx5/core/eswitch.c +4 -14
@@ -1300,20 +1300,19 @@ int mlx5_eswitch_enable_locked(struct mlx5_eswitch *esw, int num_vfs)*/ int mlx5_eswitch_enable (struct mlx5_eswitch *esw,int num_vfs) { -struct devlink *devlink;bool toggle_lag;int ret;if (!mlx5_esw_allowed (esw))return 0 ; +devl_assert_locked (priv_to_devlink (esw->dev)); + toggle_lag = !mlx5_esw_is_fdb_created (esw);if (toggle_lag)mlx5_lag_disable_change (esw->dev); - devlink =priv_to_devlink (esw->dev); -devl_lock (devlink);down_write (&esw->mode_lock);if (!mlx5_esw_is_fdb_created (esw)) { ret =mlx5_eswitch_enable_locked (esw, num_vfs);
@@ -1327,7 +1326,6 @@ int mlx5_eswitch_enable(struct mlx5_eswitch *esw, int num_vfs) esw->esw_funcs.num_vfs = num_vfs; }up_write (&esw->mode_lock); -devl_unlock (devlink);if (toggle_lag)mlx5_lag_enable_change (esw->dev);
@@ -1338,13 +1336,10 @@ int mlx5_eswitch_enable(struct mlx5_eswitch *esw, int num_vfs)/* When disabling sriov, free driver level resources. */ void mlx5_eswitch_disable_sriov (struct mlx5_eswitch *esw,bool clear_vf) { -struct devlink *devlink; -if (!mlx5_esw_allowed (esw))return ; - devlink =priv_to_devlink (esw->dev); -devl_lock (devlink); +devl_assert_locked (priv_to_devlink (esw->dev));down_write (&esw->mode_lock);/* If driver is unloaded, this function is called twice by remove_one() * and mlx5_unload(). Prevent the second call.
@@ -1373,7 +1368,6 @@ void mlx5_eswitch_disable_sriov(struct mlx5_eswitch *esw, bool clear_vf) unlock:up_write (&esw->mode_lock); -devl_unlock (devlink); }/* Free resources for corresponding eswitch mode. It is called by devlink
@@ -1407,18 +1401,14 @@ void mlx5_eswitch_disable_locked(struct mlx5_eswitch *esw)void mlx5_eswitch_disable (struct mlx5_eswitch *esw) { -struct devlink *devlink; -if (!mlx5_esw_allowed (esw))return ; +devl_assert_locked (priv_to_devlink (esw->dev));mlx5_lag_disable_change (esw->dev); - devlink =priv_to_devlink (esw->dev); -devl_lock (devlink);down_write (&esw->mode_lock);mlx5_eswitch_disable_locked (esw);up_write (&esw->mode_lock); -devl_unlock (devlink);mlx5_lag_enable_change (esw->dev); }
drivers/net/ethernet/mellanox/mlx5/core/main.c +32 -2
@@ -1304,8 +1304,10 @@ static void mlx5_unload(struct mlx5_core_dev *dev)int mlx5_init_one (struct mlx5_core_dev *dev) { +struct devlink *devlink =priv_to_devlink (dev);int err =0 ; +devl_lock (devlink);mutex_lock (&dev->intf_state_mutex); dev->state =MLX5_DEVICE_STATE_UP ;
@@ -1334,6 +1336,7 @@ int mlx5_init_one(struct mlx5_core_dev *dev) goto err_register;mutex_unlock (&dev->intf_state_mutex); +devl_unlock (devlink);return 0 ; err_register:
@@ -1348,11 +1351,15 @@ int mlx5_init_one(struct mlx5_core_dev *dev) err_function: dev->state =MLX5_DEVICE_STATE_INTERNAL_ERROR ;mutex_unlock (&dev->intf_state_mutex); +devl_unlock (devlink);return err; }void mlx5_uninit_one (struct mlx5_core_dev *dev) { +struct devlink *devlink =priv_to_devlink (dev); + +devl_lock (devlink);mutex_lock (&dev->intf_state_mutex);mlx5_unregister_device (dev);
@@ -1371,13 +1378,15 @@ void mlx5_uninit_one(struct mlx5_core_dev *dev)mlx5_function_teardown (dev, true); out:mutex_unlock (&dev->intf_state_mutex); +devl_unlock (devlink); } -int mlx5_load_one (struct mlx5_core_dev *dev,bool recovery) +int mlx5_load_one_devl_locked (struct mlx5_core_dev *dev,bool recovery) {int err =0 ;u64 timeout; +devl_assert_locked (priv_to_devlink (dev));mutex_lock (&dev->intf_state_mutex);if (test_bit (MLX5_INTERFACE_STATE_UP , &dev->intf_state)) {mlx5_core_warn (dev,"interface is up, NOP\n" );
@@ -1419,8 +1428,20 @@ int mlx5_load_one(struct mlx5_core_dev *dev, bool recovery)return err; } -void mlx5_unload_one (struct mlx5_core_dev *dev) +int mlx5_load_one (struct mlx5_core_dev *dev,bool recovery) { +struct devlink *devlink =priv_to_devlink (dev); +int ret; + +devl_lock (devlink); + ret =mlx5_load_one_devl_locked (dev, recovery); +devl_unlock (devlink); +return ret; +} + +void mlx5_unload_one_devl_locked (struct mlx5_core_dev *dev) +{ +devl_assert_locked (priv_to_devlink (dev));mutex_lock (&dev->intf_state_mutex);mlx5_detach_device (dev);
@@ -1438,6 +1459,15 @@ void mlx5_unload_one(struct mlx5_core_dev *dev)mutex_unlock (&dev->intf_state_mutex); } +void mlx5_unload_one (struct mlx5_core_dev *dev) +{ +struct devlink *devlink =priv_to_devlink (dev); + +devl_lock (devlink); +mlx5_unload_one_devl_locked (dev); +devl_unlock (devlink); +} +static const int types[] = {MLX5_CAP_GENERAL ,MLX5_CAP_GENERAL_2 ,
drivers/net/ethernet/mellanox/mlx5/core/mlx5_core.h +2 -0
@@ -290,7 +290,9 @@ void mlx5_mdev_uninit(struct mlx5_core_dev *dev);int mlx5_init_one (struct mlx5_core_dev *dev);void mlx5_uninit_one (struct mlx5_core_dev *dev);void mlx5_unload_one (struct mlx5_core_dev *dev); +void mlx5_unload_one_devl_locked (struct mlx5_core_dev *dev);int mlx5_load_one (struct mlx5_core_dev *dev,bool recovery); +int mlx5_load_one_devl_locked (struct mlx5_core_dev *dev,bool recovery);int mlx5_vport_get_other_func_cap (struct mlx5_core_dev *dev,u16 function_id,void *out);
drivers/net/ethernet/mellanox/mlx5/core/sriov.c +6 -0
@@ -154,13 +154,16 @@ mlx5_device_disable_sriov(struct mlx5_core_dev *dev, int num_vfs, bool clear_vf)static int mlx5_sriov_enable (struct pci_dev *pdev,int num_vfs) {struct mlx5_core_dev *dev =pci_get_drvdata (pdev); +struct devlink *devlink =priv_to_devlink (dev);int err; +devl_lock (devlink); err =mlx5_device_enable_sriov (dev, num_vfs);if (err) {mlx5_core_warn (dev,"mlx5_device_enable_sriov failed : %d\n" , err);return err; } +devl_unlock (devlink); err =pci_enable_sriov (pdev, num_vfs);if (err) {
@@ -173,10 +176,13 @@ static int mlx5_sriov_enable(struct pci_dev *pdev, int num_vfs)void mlx5_sriov_disable (struct pci_dev *pdev) {struct mlx5_core_dev *dev =pci_get_drvdata (pdev); +struct devlink *devlink =priv_to_devlink (dev);int num_vfs =pci_num_vf (dev->pdev);pci_disable_sriov (pdev); +devl_lock (devlink);mlx5_device_disable_sriov (dev, num_vfs, true); +devl_unlock (devlink); }int mlx5_core_sriov_configure (struct pci_dev *pdev,int num_vfs)
Signature · 1 line
2.18.2
PATCH v2 5/9 net/mlx4: Use devl_ API for devlink region create / destroy
In reply to Moshe ShemeshUse devl_ API to call devl_region_create() and devl_region_destroy() instead of devlink_region_create() and devlink_region_destroy(). Add devlink instance lock in mlx4 driver paths to these functions.
This will be used by the downstream patch to invoke mlx4 devlink reload callbacks with devlink lock held.
--- drivers/net/ethernet/mellanox/mlx4/crdump.c | 20 ++++++++++---------- drivers/net/ethernet/mellanox/mlx4/main.c | 7 +++++++ 2 files changed, 17 insertions(+), 10 deletions(-)
drivers/net/ethernet/mellanox/mlx4/crdump.c +10 -10
@@ -226,10 +226,10 @@ int mlx4_crdump_init(struct mlx4_dev *dev)/* Create cr-space region */ crdump->region_crspace = -devlink_region_create (devlink, - ®ion_cr_space_ops, -MAX_NUM_OF_DUMPS_TO_STORE , -pci_resource_len (pdev,0 )); +devl_region_create (devlink, + ®ion_cr_space_ops, +MAX_NUM_OF_DUMPS_TO_STORE , +pci_resource_len (pdev,0 ));if (IS_ERR (crdump->region_crspace)) mlx4_warn(dev, "crdump: create devlink region %s err %ld\n", region_cr_space_str,
@@ -237,10 +237,10 @@ int mlx4_crdump_init(struct mlx4_dev *dev)/* Create fw-health region */ crdump->region_fw_health = -devlink_region_create (devlink, - ®ion_fw_health_ops, -MAX_NUM_OF_DUMPS_TO_STORE , -HEALTH_BUFFER_SIZE ); +devl_region_create (devlink, + ®ion_fw_health_ops, +MAX_NUM_OF_DUMPS_TO_STORE , +HEALTH_BUFFER_SIZE );if (IS_ERR (crdump->region_fw_health)) mlx4_warn(dev, "crdump: create devlink region %s err %ld\n", region_fw_health_str,
@@ -253,6 +253,6 @@ void mlx4_crdump_end(struct mlx4_dev *dev) {struct mlx4_fw_crdump *crdump = &dev->persist->crdump; -devlink_region_destroy (crdump->region_fw_health); -devlink_region_destroy (crdump->region_crspace); +devl_region_destroy (crdump->region_fw_health); +devl_region_destroy (crdump->region_crspace); }
drivers/net/ethernet/mellanox/mlx4/main.c +7 -0
@@ -3732,6 +3732,7 @@ static int __mlx4_init_one(struct pci_dev *pdev, int pci_dev_data,int prb_vf[MLX4_MAX_PORTS +1 ] = {0 ,0 ,0 };const int param_map[MLX4_MAX_PORTS +1 ][MLX4_MAX_PORTS +1 ] = { {2 ,0 ,0 }, {0 ,1 ,2 }, {0 ,1 ,2 } }; +struct devlink *devlink =priv_to_devlink (priv);unsigned total_vfs =0 ;unsigned int i;
@@ -3844,7 +3845,9 @@ static int __mlx4_init_one(struct pci_dev *pdev, int pci_dev_data, } } +devl_lock (devlink); err =mlx4_crdump_init (&priv->dev); +devl_unlock (devlink);if (err) goto err_release_regions;
@@ -3862,7 +3865,9 @@ static int __mlx4_init_one(struct pci_dev *pdev, int pci_dev_data,mlx4_catas_end (&priv->dev); err_crdump: +devl_lock (devlink);mlx4_crdump_end (&priv->dev); +devl_unlock (devlink); err_release_regions:pci_release_regions (pdev);
@@ -4161,7 +4166,9 @@ static void mlx4_remove_one(struct pci_dev *pdev)else mlx4_info (dev,"%s: interface is down\n" , __func__);mlx4_catas_end (dev); +devl_lock (devlink);mlx4_crdump_end (dev); +devl_unlock (devlink);if (dev->flags &MLX4_FLAG_SRIOV && !active_vfs) {mlx4_warn (dev,"Disabling SR-IOV\n" );pci_disable_sriov (pdev);
Signature · 1 line
2.18.2
PATCH v2 6/9 net/mlx4: Use devl_ API for devlink port register / unregister
In reply to Moshe ShemeshUse devl_ API to call devl_port_register() and devl_port_unregister() instead of devlink_port_register() and devlink_port_unregister(). Add devlink instance lock in mlx4 driver paths to these functions.
This will be used by the downstream patch to invoke mlx4 devlink reload callbacks with devlink lock held.
--- drivers/net/ethernet/mellanox/mlx4/main.c | 16 ++++++++++++---- 1 file changed, 12 insertions(+), 4 deletions(-)
drivers/net/ethernet/mellanox/mlx4/main.c +12 -4
@@ -3033,7 +3033,7 @@ static int mlx4_init_port_info(struct mlx4_dev *dev, int port)struct mlx4_port_info *info = &mlx4_priv (dev)->port[port];int err; - err =devlink_port_register (devlink, &info->devlink_port, port); + err =devl_port_register (devlink, &info->devlink_port, port);if (err)return err;
@@ -3071,7 +3071,7 @@ static int mlx4_init_port_info(struct mlx4_dev *dev, int port) err =device_create_file (&dev->persist->pdev->dev, &info->port_attr);if (err) {mlx4_err (dev,"Failed to create file for port %d\n" , port); -devlink_port_unregister (&info->devlink_port); +devl_port_unregister (&info->devlink_port); info->port =-1 ;return err; }
@@ -3093,7 +3093,7 @@ static int mlx4_init_port_info(struct mlx4_dev *dev, int port)mlx4_err (dev,"Failed to create mtu file for port %d\n" , port);device_remove_file (&info->dev->persist->pdev->dev, &info->port_attr); -devlink_port_unregister (&info->devlink_port); +devl_port_unregister (&info->devlink_port); info->port =-1 ;return err; }
@@ -3109,7 +3109,7 @@ static void mlx4_cleanup_port_info(struct mlx4_port_info *info)device_remove_file (&info->dev->persist->pdev->dev, &info->port_attr);device_remove_file (&info->dev->persist->pdev->dev, &info->port_mtu_attr); -devlink_port_unregister (&info->devlink_port); +devl_port_unregister (&info->devlink_port);#ifdef CONFIG_RFS_ACCEL free_irq_cpu_rmap (info->rmap);
@@ -3333,6 +3333,7 @@ static int mlx4_load_one(struct pci_dev *pdev, int pci_dev_data, int total_vfs, int *nvfs, struct mlx4_priv *priv, int reset_flow) { + struct devlink *devlink = priv_to_devlink(priv);struct mlx4_dev *dev;unsigned sum =0 ;int err;
@@ -3629,6 +3630,7 @@ static int mlx4_load_one(struct pci_dev *pdev, int pci_dev_data, } } +devl_lock (devlink);for (port =1 ; port <= dev->caps.num_ports; port++) { err =mlx4_init_port_info (dev, port);if (err)
@@ -3642,6 +3644,7 @@ static int mlx4_load_one(struct pci_dev *pdev, int pci_dev_data,if (err) goto err_port; +devl_unlock (devlink);mlx4_request_modules (dev);mlx4_sense_init (dev);
@@ -3658,6 +3661,7 @@ static int mlx4_load_one(struct pci_dev *pdev, int pci_dev_data, err_port:for (--port; port >=1 ; --port)mlx4_cleanup_port_info (&priv->port[port]); +devl_unlock (devlink);mlx4_cleanup_default_counters (dev);if (!mlx4_is_slave (dev))
@@ -4061,8 +4065,10 @@ static void mlx4_unload_one(struct pci_dev *pdev)struct mlx4_dev *dev = persist->dev;struct mlx4_priv *priv =mlx4_priv (dev);int pci_dev_data; +struct devlink *devlink;int p, i; + devlink =priv_to_devlink (priv);if (priv->removed)return ;
@@ -4078,10 +4084,12 @@ static void mlx4_unload_one(struct pci_dev *pdev)mlx4_stop_sense (dev);mlx4_unregister_device (dev); +devl_lock (devlink);for (p =1 ; p <= dev->caps.num_ports; p++) {mlx4_cleanup_port_info (&priv->port[p]);mlx4_CLOSE_PORT (dev, p); } +devl_unlock (devlink);if (mlx4_is_master (dev)) mlx4_free_resource_tracker(dev,
Signature · 1 line
2.18.2
PATCH v2 7/9 net/mlx4: Lock mlx4 devlink reload callback
In reply to Moshe ShemeshChange devlink instance locks in mlx4 driver to have devlink reload callback locked, while keeping all driver paths which leads to devl_ API functions called by the mlx4 driver locked.
--- drivers/net/ethernet/mellanox/mlx4/catas.c | 5 +++ drivers/net/ethernet/mellanox/mlx4/main.c | 45 +++++++++++++++------- 2 files changed, 37 insertions(+), 13 deletions(-)
drivers/net/ethernet/mellanox/mlx4/catas.c +5 -0
@@ -204,9 +204,13 @@ void mlx4_enter_error_state(struct mlx4_dev_persistent *persist)static void mlx4_handle_error_state (struct mlx4_dev_persistent *persist) { +struct mlx4_dev *dev = persist->dev; +struct devlink *devlink;int err =0 ;mlx4_enter_error_state (persist); + devlink =priv_to_devlink (mlx4_priv (dev)); +devl_lock (devlink);mutex_lock (&persist->interface_state_mutex);if (persist->interface_state &MLX4_INTERFACE_STATE_UP && !(persist->interface_state &MLX4_INTERFACE_STATE_DELETION )) {
@@ -215,6 +219,7 @@ static void mlx4_handle_error_state(struct mlx4_dev_persistent *persist) err); }mutex_unlock (&persist->interface_state_mutex); +devl_unlock (devlink); }static void dump_err_buf (struct mlx4_dev *dev)
drivers/net/ethernet/mellanox/mlx4/main.c +32 -13
234 lines not on this page · raw message
Signature · 1 line
2.18.2
PATCH v2 8/9 net/mlx5: Lock mlx5 devlink health recovery callback
In reply to Moshe ShemeshChange devlink instance locks in mlx5 driver to have devlink health recovery callback locked, while keeping all driver paths which lead to devl_ API functions called by the driver locked.
--- drivers/net/ethernet/mellanox/mlx5/core/health.c | 12 +++++++++++- drivers/net/ethernet/mellanox/mlx5/core/main.c | 4 ++-- 2 files changed, 13 insertions(+), 3 deletions(-)
drivers/net/ethernet/mellanox/mlx5/core/health.c +11 -1
@@ -622,8 +622,14 @@ mlx5_fw_fatal_reporter_recover(struct devlink_health_reporter *reporter, struct netlink_ext_ack *extack) {struct mlx5_core_dev *dev =devlink_health_reporter_priv (reporter); +struct devlink *devlink =priv_to_devlink (dev); +int ret; -return mlx5_health_try_recover (dev); +devl_lock (devlink); + ret =mlx5_health_try_recover (dev); +devl_unlock (devlink); + +return ret; } static int
@@ -666,16 +672,20 @@ static void mlx5_fw_fatal_reporter_err_work(struct work_struct *work)struct mlx5_fw_reporter_ctx fw_reporter_ctx;struct mlx5_core_health *health;struct mlx5_core_dev *dev; +struct devlink *devlink;struct mlx5_priv *priv; health = container_of(work, struct mlx5_core_health, fatal_report_work); priv = container_of(health, struct mlx5_priv, health); dev = container_of(priv, struct mlx5_core_dev, priv); + devlink =priv_to_devlink (dev);enter_error_state (dev, false);if (IS_ERR_OR_NULL (health->fw_fatal_reporter)) { +devl_lock (devlink);if (mlx5_health_try_recover (dev))mlx5_core_err (dev,"health recovery failed\n" ); +devl_unlock (devlink);return ; } fw_reporter_ctx.err_synd = health->synd;
drivers/net/ethernet/mellanox/mlx5/core/main.c +2 -2
@@ -1932,7 +1932,7 @@ MODULE_DEVICE_TABLE(pci, mlx5_core_pci_table);void mlx5_disable_device (struct mlx5_core_dev *dev) {mlx5_error_sw_reset (dev); -mlx5_unload_one (dev); +mlx5_unload_one_devl_locked (dev); }int mlx5_recover_device (struct mlx5_core_dev *dev)
@@ -1943,7 +1943,7 @@ int mlx5_recover_device(struct mlx5_core_dev *dev)return -EIO ; } -return mlx5_load_one (dev, true); +return mlx5_load_one_devl_locked (dev, true); }static struct pci_driver mlx5_core_driver = {
Signature · 1 line
2.18.2
PATCH v2 9/9 devlink: Hold the instance lock in health callbacks
In reply to Moshe ShemeshLet the core take the devlink instance lock around health callbacks and remove the now redundant locking in the drivers.
---v1->v2:
- added lock while calling devlink_health_reporter_recover() from devlink core (bug found by test)
--- .../net/ethernet/mellanox/mlx5/core/health.c | 8 +---- net/core/devlink.c | 30 +++++++++---------- 2 files changed, 15 insertions(+), 23 deletions(-)
drivers/net/ethernet/mellanox/mlx5/core/health.c +1 -7
@@ -622,14 +622,8 @@ mlx5_fw_fatal_reporter_recover(struct devlink_health_reporter *reporter, struct netlink_ext_ack *extack) {struct mlx5_core_dev *dev =devlink_health_reporter_priv (reporter); -struct devlink *devlink =priv_to_devlink (dev); -int ret; -devl_lock (devlink); - ret =mlx5_health_try_recover (dev); -devl_unlock (devlink); - -return ret; +return mlx5_health_try_recover (dev); } static int
net/core/devlink.c +14 -16
@@ -7553,6 +7553,7 @@ int devlink_health_report(struct devlink_health_reporter *reporter,enum devlink_health_reporter_state prev_health_state;struct devlink *devlink = reporter->devlink;unsigned long recover_ts_threshold; +int ret;/* write a log message of the current error */ WARN_ON (!msg);
@@ -7586,11 +7587,14 @@ int devlink_health_report(struct devlink_health_reporter *reporter,mutex_unlock (&reporter->dump_lock); } -if (reporter->auto_recover) -return devlink_health_reporter_recover (reporter, - priv_ctx,NULL ); +if (!reporter->auto_recover) +return 0 ; -return 0 ; +devl_lock (devlink); + ret =devlink_health_reporter_recover (reporter, priv_ctx,NULL ); +devl_unlock (devlink); + +return ret; }EXPORT_SYMBOL_GPL (devlink_health_report);
@@ -9264,8 +9268,7 @@ static const struct genl_small_ops devlink_nl_ops[] = { .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP, .doit = devlink_nl_cmd_health_reporter_get_doit, .dumpit = devlink_nl_cmd_health_reporter_get_dumpit, - .internal_flags =DEVLINK_NL_FLAG_NEED_DEVLINK_OR_PORT | - DEVLINK_NL_FLAG_NO_LOCK, + .internal_flags = DEVLINK_NL_FLAG_NEED_DEVLINK_OR_PORT, /* can be retrieved by unprivileged users */ }, {
@@ -9273,24 +9276,21 @@ static const struct genl_small_ops devlink_nl_ops[] = { .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP, .doit = devlink_nl_cmd_health_reporter_set_doit, .flags = GENL_ADMIN_PERM, - .internal_flags = DEVLINK_NL_FLAG_NEED_DEVLINK_OR_PORT | - DEVLINK_NL_FLAG_NO_LOCK, + .internal_flags = DEVLINK_NL_FLAG_NEED_DEVLINK_OR_PORT, }, { .cmd = DEVLINK_CMD_HEALTH_REPORTER_RECOVER, .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP, .doit = devlink_nl_cmd_health_reporter_recover_doit, .flags = GENL_ADMIN_PERM, - .internal_flags =DEVLINK_NL_FLAG_NEED_DEVLINK_OR_PORT | - DEVLINK_NL_FLAG_NO_LOCK, + .internal_flags = DEVLINK_NL_FLAG_NEED_DEVLINK_OR_PORT, }, { .cmd = DEVLINK_CMD_HEALTH_REPORTER_DIAGNOSE, .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP, .doit = devlink_nl_cmd_health_reporter_diagnose_doit, .flags = GENL_ADMIN_PERM, - .internal_flags =DEVLINK_NL_FLAG_NEED_DEVLINK_OR_PORT | - DEVLINK_NL_FLAG_NO_LOCK, + .internal_flags = DEVLINK_NL_FLAG_NEED_DEVLINK_OR_PORT, }, { .cmd = DEVLINK_CMD_HEALTH_REPORTER_DUMP_GET,
@@ -9304,16 +9304,14 @@ static const struct genl_small_ops devlink_nl_ops[] = { .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP, .doit = devlink_nl_cmd_health_reporter_dump_clear_doit, .flags = GENL_ADMIN_PERM, - .internal_flags = DEVLINK_NL_FLAG_NEED_DEVLINK_OR_PORT | - DEVLINK_NL_FLAG_NO_LOCK, + .internal_flags = DEVLINK_NL_FLAG_NEED_DEVLINK_OR_PORT, }, { .cmd = DEVLINK_CMD_HEALTH_REPORTER_TEST, .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP, .doit = devlink_nl_cmd_health_reporter_test_doit, .flags = GENL_ADMIN_PERM, - .internal_flags =DEVLINK_NL_FLAG_NEED_DEVLINK_OR_PORT | - DEVLINK_NL_FLAG_NO_LOCK, + .internal_flags = DEVLINK_NL_FLAG_NEED_DEVLINK_OR_PORT, }, { .cmd = DEVLINK_CMD_FLASH_UPDATE,
Signature · 1 line
2.18.2
This series was applied to netdev/net-next.git (master) by Jakub Kicinski <[email protected]>:
Hello:
This series was applied to netdev/net-next.git (master) by Jakub Kicinski <[email protected]>:
Quoted 11 lines · On Thu, 28 Jul 2022 18:53:41 +0300 you wrote:
Prepare mlx4 and mlx5 drivers to have all devlink callbacks called with devlink instance locked. Change mlx4 driver to use
devl_API where needed to have devlink reload callbacks locked. Change mlx5 driver to usedevl_API where needed to have devlink reload and devlink health callbacks locked.As mlx5 is the only driver which needed changes to enable calling health callbacks with devlink instance locked, this patchset also removes
DEVLINK_NL_FLAG_NO_LOCKflag from devlink health callbacks.[...]
Here is the summary with links:
- [net-next,v2,1/9] net: devlink: remove region snapshot ID tracking dependency on devlink->lock
https://git.kernel.org/netdev/net-next/c/5502e8712c9b - [net-next,v2,2/9] net: devlink: remove region snapshots list dependency on devlink->lock https://git.kernel.org/netdev/net-next/c/2dec18ad826f - [net-next,v2,3/9] net/mlx5: Move fw reset unload to mlx5_fw_reset_complete_reload https://git.kernel.org/netdev/net-next/c/c12f4c6ac3b4 - [net-next,v2,4/9] net/mlx5: Lock mlx5 devlink reload callbacks https://git.kernel.org/netdev/net-next/c/84a433a40d0e - [net-next,v2,5/9] net/mlx4: Use devl_ API for devlink region create / destroy https://git.kernel.org/netdev/net-next/c/9cb7e94a78b5 - [net-next,v2,6/9] net/mlx4: Use devl_ API for devlink port register / unregister https://git.kernel.org/netdev/net-next/c/a8c05514b2f8 - [net-next,v2,7/9] net/mlx4: Lock mlx4 devlink reload callback https://git.kernel.org/netdev/net-next/c/60d7ceea4b2a - [net-next,v2,8/9] net/mlx5: Lock mlx5 devlink health recovery callback https://git.kernel.org/netdev/net-next/c/d3dbdc9f8ddc - [net-next,v2,9/9] devlink: Hold the instance lock in health callbacks https://git.kernel.org/netdev/net-next/c/c90005b5f75c
You are awesome, thank you!
Signature · 2 lines
Deet-doot-dot, I am a bot. https://korg.docs.kernel.org/patchwork/pwbot.html