Thread (11 messages) flat view 11 messages, 3 authors, 2021-08-06

Re: [PATCH net-next v1] netdevsim: Forbid devlink reload when adding or deleting ports

From: Leon Romanovsky <leon@kernel.org>
Date: 2021-08-05 14:33:44
Also in: lkml

On Thu, Aug 05, 2021 at 07:23:42AM -0700, Jakub Kicinski wrote:
On Thu, 5 Aug 2021 16:51:31 +0300 Leon Romanovsky wrote:
quoted
quoted
quoted
+	nsim_bus_dev = nsim_dev->nsim_bus_dev;
+	if (!mutex_trylock(&nsim_bus_dev->nsim_bus_reload_lock))
+		return -EOPNOTSUPP;  
Why not -EBUSY?  
This is what devlink_reload_disable() returns, so I kept same error.
It is not important at all.

What about the following change on top of this patch?
LGTM, the only question is whether we should leave in_reload true 
if nsim_dev->fail_reload is set.
I don't think so, it will block add/delete ports.
quoted
@@ -889,17 +890,26 @@ static int nsim_dev_reload_up(struct devlink *devlink, enum devlink_reload_actio
 			      struct netlink_ext_ack *extack)
 {
 	struct nsim_dev *nsim_dev = devlink_priv(devlink);
+	struct nsim_bus_dev *nsim_bus_dev;
+	int ret;
+
+	nsim_bus_dev = nsim_dev->nsim_bus_dev;
+	mutex_lock(&nsim_bus_dev->nsim_bus_reload_lock);
+	nsim_bus_dev->in_reload = false;
 
 	if (nsim_dev->fail_reload) {
 		/* For testing purposes, user set debugfs fail_reload
 		 * value to true. Fail right away.
 		 */
 		NL_SET_ERR_MSG_MOD(extack, "User setup the reload to fail for testing purposes");
+		mutex_unlock(&nsim_bus_dev->nsim_bus_reload_lock);
 		return -EINVAL;
 	}
 
 	*actions_performed = BIT(DEVLINK_RELOAD_ACTION_DRIVER_REINIT);
-	return nsim_dev_reload_create(nsim_dev, extack);
+	ret = nsim_dev_reload_create(nsim_dev, extack);
+	mutex_unlock(&nsim_bus_dev->nsim_bus_reload_lock);
+	return ret;
 }
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help