[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <Zr4GxjRf5la7wBXM@gmail.com>
Date: Thu, 15 Aug 2024 06:46:46 -0700
From: Breno Leitao <leitao@...ian.org>
To: Paolo Abeni <pabeni@...hat.com>
Cc: kuba@...nel.org, davem@...emloft.net, edumazet@...gle.com,
thepacketgeek@...il.com, netdev@...r.kernel.org,
linux-kernel@...r.kernel.org, Aijay Adams <aijay@...a.com>
Subject: Re: [PATCH net-next] net: netconsole: Populate dynamic entry even if
netpoll fails
On Wed, Aug 14, 2024 at 12:06:48PM +0200, Paolo Abeni wrote:
> I fear the late cleanup could still be dangerous - what if multiple,
> consecutive, enabled_store() on the same target fails?
>
> I *think* it would be safer always zeroing np->dev in the error path of
> netpoll_setup().
>
> It could be a separate patch for bisectability.
>
> Side note: I additionally think that in the same error path we should
> conditionally clear np->local_ip.ip, if the previous code initialized such
> field, or we could get weird results if e.g.
> - a target uses eth0 with local_ip == 0
> - enabled_store() of such target fails e.g. due ndo_netpoll_setup() failure
> - address on eth0 changes for some reason
> - anoter enabled_store() is issued on the same target.
>
> At this point the netpoll target should be wrongly using the old address.
Agree with you. I think we always want to keep struct netpoll objects
either initialized or unitialized, not keeping them half-baked.
How about the following patch:
netpoll: Ensure clean state on setup failures
Modify netpoll_setup() and __netpoll_setup() to ensure that the netpoll
structure (np) is left in a clean state if setup fails for any reason.
This prevents carrying over misconfigured fields in case of partial
setup success.
Key changes:
- np->dev is now set only after successful setup, ensuring it's always
NULL if netpoll is not configured or if netpoll_setup() fails.
- np->local_ip is zeroed if netpoll setup doesn't complete successfully.
- Added DEBUG_NET_WARN_ON_ONCE() checks to catch unexpected states.
These changes improve the reliability of netpoll configuration, since it
assures that the structure is fully initialized or totally unset.
Suggested-by: Paolo Abeni <pabeni@...hat.com>
Signed-off-by: Breno Leitao <leitao@...ian.org>
diff --git a/net/core/netpoll.c b/net/core/netpoll.c
index a58ea724790c..348d76a51c20 100644
--- a/net/core/netpoll.c
+++ b/net/core/netpoll.c
@@ -626,12 +626,10 @@ int __netpoll_setup(struct netpoll *np, struct net_device *ndev)
const struct net_device_ops *ops;
int err;
- np->dev = ndev;
- strscpy(np->dev_name, ndev->name, IFNAMSIZ);
-
+ DEBUG_NET_WARN_ON_ONCE(np->dev);
if (ndev->priv_flags & IFF_DISABLE_NETPOLL) {
np_err(np, "%s doesn't support polling, aborting\n",
- np->dev_name);
+ ndev->name);
err = -ENOTSUPP;
goto out;
}
@@ -649,7 +647,7 @@ int __netpoll_setup(struct netpoll *np, struct net_device *ndev)
refcount_set(&npinfo->refcnt, 1);
- ops = np->dev->netdev_ops;
+ ops = ndev->netdev_ops;
if (ops->ndo_netpoll_setup) {
err = ops->ndo_netpoll_setup(ndev, npinfo);
if (err)
@@ -660,6 +658,8 @@ int __netpoll_setup(struct netpoll *np, struct net_device *ndev)
refcount_inc(&npinfo->refcnt);
}
+ np->dev = ndev;
+ strscpy(np->dev_name, ndev->name, IFNAMSIZ);
npinfo->netpoll = np;
/* last thing to do is link it to the net device structure */
@@ -681,6 +681,7 @@ int netpoll_setup(struct netpoll *np)
int err;
rtnl_lock();
+ DEBUG_NET_WARN_ON_ONCE(np->dev);
if (np->dev_name[0]) {
struct net *net = current->nsproxy->net_ns;
ndev = __dev_get_by_name(net, np->dev_name);
@@ -782,11 +783,14 @@ int netpoll_setup(struct netpoll *np)
err = __netpoll_setup(np, ndev);
if (err)
- goto put;
+ goto clear_ip;
rtnl_unlock();
return 0;
+clear_ip:
+ memset(&np->local_ip, 0, sizeof(np->local_ip));
put:
+ DEBUG_NET_WARN_ON_ONCE(np->dev);
netdev_put(ndev, &np->dev_tracker);
unlock:
rtnl_unlock();
Powered by blists - more mailing lists