When we terminated the dump, the callback isn't running, so cb_running
should be set to false to be logically consistent.

Fixes: 1904fb9ebf91 ("netlink: terminate outstanding dump on socket close")
Fixes: 16b304f3404f ("netlink: Eliminate kmalloc in netlink dump operation.")
Signed-off-by: Siddh Raman Pant <siddh.raman.pant@oracle.com>
---
I found this by inspection and was thinking why it isn't being done. So
I thought I should ask by sending a patch.

 net/netlink/af_netlink.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/net/netlink/af_netlink.c b/net/netlink/af_netlink.c
index 85311226183a..f8f13058a46e 100644
--- a/net/netlink/af_netlink.c
+++ b/net/netlink/af_netlink.c
@@ -771,6 +771,7 @@ static int netlink_release(struct socket *sock)
 			nlk->cb.done(&nlk->cb);
 		module_put(nlk->cb.module);
 		kfree_skb(nlk->cb.skb);
+		WRITE_ONCE(nlk->cb_running, false);
 	}
 
 	module_put(nlk->module);
-- 
2.45.2

On Fri, 14 Feb 2025 12:28:49 +0530 Siddh Raman Pant wrote:
> When we terminated the dump, the callback isn't running, so cb_running
> should be set to false to be logically consistent.
> 
> Fixes: 1904fb9ebf91 ("netlink: terminate outstanding dump on socket close")
> Fixes: 16b304f3404f ("netlink: Eliminate kmalloc in netlink dump operation.")
> Signed-off-by: Siddh Raman Pant <siddh.raman.pant@oracle.com>
> ---
> I found this by inspection and was thinking why it isn't being done. So
> I thought I should ask by sending a patch.

I only see uses in places called by user space. So code which can't 
be run once we're in release.

You can send as a cleanup, if you want, but with more analysis in
the commit msg, and no fixes tags.

When we terminated the dump, the callback isn't running, so cb_running
should be set to false to be logically consistent.

cb_running signifies whether a dump is ongoing. It is set to true in
cb->start(), and is checked in netlink_dump() to be true initially.
After the dump, it is set to false in the same function.

When we terminate a dump before it ends (see 1904fb9ebf91 ("netlink:
terminate outstanding dump on socket close")), we do not unset the
boolean flag cb_running. This is wrong - since the dump is no longer
running (we terminated it), it must be set to false to convey the
correct state.

Signed-off-by: Siddh Raman Pant <siddh.raman.pant@oracle.com>
---
 net/netlink/af_netlink.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/net/netlink/af_netlink.c b/net/netlink/af_netlink.c
index 85311226183a..f8f13058a46e 100644
--- a/net/netlink/af_netlink.c
+++ b/net/netlink/af_netlink.c
@@ -771,6 +771,7 @@ static int netlink_release(struct socket *sock)
 			nlk->cb.done(&nlk->cb);
 		module_put(nlk->cb.module);
 		kfree_skb(nlk->cb.skb);
+		WRITE_ONCE(nlk->cb_running, false);
 	}
 
 	module_put(nlk->module);
-- 
2.45.2