net/mlx5: Delay events till ib registration ends
When mlx5_ib registers itself to mlx5_core as an interface, it will
call mlx5_add_device which will call mlx5_ib interface add callback,
in case the latter successfully returns, only then mlx5_core will add
it to the interface list and async events will be forwarded to mlx5_ib.
Between mlx5_ib interface add callback and mlx5_core adding the mlx5_ib
interface to its devices list, arriving mlx5_core events can be missed
by the new mlx5_ib registering interface.
In other words:
thread 1: mlx5_ib: mlx5_register_interface(dev)
thread 1: mlx5_core: mlx5_add_device(dev)
thread 1: mlx5_core: ctx = dev->add => (mlx5_ib)->mlx5_ib_add
thread 2: mlx5_core_event: **new event arrives, forward to dev_list
thread 1: mlx5_core: add_ctx_to_dev_list(ctx)
/* previous event was missed by the new interface.*/
It is ok to miss events before dev->add (mlx5_ib)->mlx5_ib_add_device
but not after.
We fix this race by accumulating the events that come between the
ib_register_device (inside mlx5_add_device->(dev->add)) till the adding
to the list completes and fire them to the new registering interface
after that.
Fixes: f1ee87fe55
("net/mlx5: Organize device list API in one place")
Signed-off-by: Erez Shitrit <erezsh@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
This commit is contained in:
Родитель
e80541ecab
Коммит
97834eba7c
|
@ -45,11 +45,70 @@ struct mlx5_device_context {
|
|||
unsigned long state;
|
||||
};
|
||||
|
||||
struct mlx5_delayed_event {
|
||||
struct list_head list;
|
||||
struct mlx5_core_dev *dev;
|
||||
enum mlx5_dev_event event;
|
||||
unsigned long param;
|
||||
};
|
||||
|
||||
enum {
|
||||
MLX5_INTERFACE_ADDED,
|
||||
MLX5_INTERFACE_ATTACHED,
|
||||
};
|
||||
|
||||
static void add_delayed_event(struct mlx5_priv *priv,
|
||||
struct mlx5_core_dev *dev,
|
||||
enum mlx5_dev_event event,
|
||||
unsigned long param)
|
||||
{
|
||||
struct mlx5_delayed_event *delayed_event;
|
||||
|
||||
delayed_event = kzalloc(sizeof(*delayed_event), GFP_ATOMIC);
|
||||
if (!delayed_event) {
|
||||
mlx5_core_err(dev, "event %d is missed\n", event);
|
||||
return;
|
||||
}
|
||||
|
||||
mlx5_core_dbg(dev, "Accumulating event %d\n", event);
|
||||
delayed_event->dev = dev;
|
||||
delayed_event->event = event;
|
||||
delayed_event->param = param;
|
||||
list_add_tail(&delayed_event->list, &priv->waiting_events_list);
|
||||
}
|
||||
|
||||
static void fire_delayed_event_locked(struct mlx5_device_context *dev_ctx,
|
||||
struct mlx5_core_dev *dev,
|
||||
struct mlx5_priv *priv)
|
||||
{
|
||||
struct mlx5_delayed_event *de;
|
||||
struct mlx5_delayed_event *n;
|
||||
|
||||
/* stop delaying events */
|
||||
priv->is_accum_events = false;
|
||||
|
||||
/* fire all accumulated events before new event comes */
|
||||
list_for_each_entry_safe(de, n, &priv->waiting_events_list, list) {
|
||||
dev_ctx->intf->event(dev, dev_ctx->context, de->event, de->param);
|
||||
list_del(&de->list);
|
||||
kfree(de);
|
||||
}
|
||||
}
|
||||
|
||||
static void cleanup_delayed_evets(struct mlx5_priv *priv)
|
||||
{
|
||||
struct mlx5_delayed_event *de;
|
||||
struct mlx5_delayed_event *n;
|
||||
|
||||
spin_lock_irq(&priv->ctx_lock);
|
||||
priv->is_accum_events = false;
|
||||
list_for_each_entry_safe(de, n, &priv->waiting_events_list, list) {
|
||||
list_del(&de->list);
|
||||
kfree(de);
|
||||
}
|
||||
spin_unlock_irq(&priv->ctx_lock);
|
||||
}
|
||||
|
||||
void mlx5_add_device(struct mlx5_interface *intf, struct mlx5_priv *priv)
|
||||
{
|
||||
struct mlx5_device_context *dev_ctx;
|
||||
|
@ -63,6 +122,12 @@ void mlx5_add_device(struct mlx5_interface *intf, struct mlx5_priv *priv)
|
|||
return;
|
||||
|
||||
dev_ctx->intf = intf;
|
||||
/* accumulating events that can come after mlx5_ib calls to
|
||||
* ib_register_device, till adding that interface to the events list.
|
||||
*/
|
||||
|
||||
priv->is_accum_events = true;
|
||||
|
||||
dev_ctx->context = intf->add(dev);
|
||||
set_bit(MLX5_INTERFACE_ADDED, &dev_ctx->state);
|
||||
if (intf->attach)
|
||||
|
@ -71,6 +136,9 @@ void mlx5_add_device(struct mlx5_interface *intf, struct mlx5_priv *priv)
|
|||
if (dev_ctx->context) {
|
||||
spin_lock_irq(&priv->ctx_lock);
|
||||
list_add_tail(&dev_ctx->list, &priv->ctx_list);
|
||||
|
||||
fire_delayed_event_locked(dev_ctx, dev, priv);
|
||||
|
||||
#ifdef CONFIG_INFINIBAND_ON_DEMAND_PAGING
|
||||
if (dev_ctx->intf->pfault) {
|
||||
if (priv->pfault) {
|
||||
|
@ -84,6 +152,8 @@ void mlx5_add_device(struct mlx5_interface *intf, struct mlx5_priv *priv)
|
|||
spin_unlock_irq(&priv->ctx_lock);
|
||||
} else {
|
||||
kfree(dev_ctx);
|
||||
/* delete all accumulated events */
|
||||
cleanup_delayed_evets(priv);
|
||||
}
|
||||
}
|
||||
|
||||
|
@ -341,6 +411,9 @@ void mlx5_core_event(struct mlx5_core_dev *dev, enum mlx5_dev_event event,
|
|||
|
||||
spin_lock_irqsave(&priv->ctx_lock, flags);
|
||||
|
||||
if (priv->is_accum_events)
|
||||
add_delayed_event(priv, dev, event, param);
|
||||
|
||||
list_for_each_entry(dev_ctx, &priv->ctx_list, list)
|
||||
if (dev_ctx->intf->event)
|
||||
dev_ctx->intf->event(dev, dev_ctx->context, event, param);
|
||||
|
|
|
@ -1343,6 +1343,9 @@ static int init_one(struct pci_dev *pdev,
|
|||
mutex_init(&dev->pci_status_mutex);
|
||||
mutex_init(&dev->intf_state_mutex);
|
||||
|
||||
INIT_LIST_HEAD(&priv->waiting_events_list);
|
||||
priv->is_accum_events = false;
|
||||
|
||||
#ifdef CONFIG_INFINIBAND_ON_DEMAND_PAGING
|
||||
err = init_srcu_struct(&priv->pfault_srcu);
|
||||
if (err) {
|
||||
|
|
|
@ -647,6 +647,9 @@ struct mlx5_priv {
|
|||
struct list_head ctx_list;
|
||||
spinlock_t ctx_lock;
|
||||
|
||||
struct list_head waiting_events_list;
|
||||
bool is_accum_events;
|
||||
|
||||
struct mlx5_flow_steering *steering;
|
||||
struct mlx5_mpfs *mpfs;
|
||||
struct mlx5_eswitch *eswitch;
|
||||
|
|
Загрузка…
Ссылка в новой задаче