* [PATCH net] net: dcb: flush rewrite table on device unregister
@ 2026-11-11 3:31 Wenbo Sun
2026-11-11 8:54 ` Daniel Machon
2026-11-11 13:01 ` Petr Machata
0 siblings, 2 replies; 3+ messages in thread
From: Wenbo Sun @ 2026-11-11 3:31 UTC (permalink / raw)
To: Daniel Machon, Petr Machata, David S . Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni, Simon Horman
Cc: Alexei Lazar, netdev, linux-kernel, sunwenbo, 1342116488
In dcbnl_flush_dev(), which is called upon NETDEV_UNREGISTER for net
devices with dcbnl_ops, only entries in dcb_app_list matching
dev->ifindex are freed. The entries in dcb_rewr_list added via
dcb_setrewr() are not flushed.
This causes a memory leak of struct dcb_app_type nodes in dcb_rewr_list
when a DCB-capable net device is unregistered, and leaves stale entries
in the global rewrite table that could inappropriately match if another
net device later reuses the same ifindex.
Iterate over dcb_rewr_list in dcbnl_flush_dev() and remove and free all
entries matching dev->ifindex, mirroring the cleanup performed for
dcb_app_list.
This issue was found via LLM-assisted code review, confirmed by
reproducing the leaked rewrite table entry upon netdevice unregistration
in a User-Mode Linux (ARCH=um) environment, and verified in UML by
confirming that all rewrite table entries for the unregistered netdevice
are flushed.
Fixes: 622f1b2fae2e ("net: dcb: add new rewrite table")
Signed-off-by: Wenbo Sun <sunwenbo@cmss.chinamobile.com>
---
net/dcb/dcbnl.c | 7 +++++++
1 file changed, 7 insertions(+)
diff --git a/net/dcb/dcbnl.c b/net/dcb/dcbnl.c
index ad0aebe0ad3d..95250658ecc1 100644
--- a/net/dcb/dcbnl.c
+++ b/net/dcb/dcbnl.c
@@ -2393,6 +2393,13 @@ static void dcbnl_flush_dev(struct net_device *dev)
}
}
+ list_for_each_entry_safe(itr, tmp, &dcb_rewr_list, list) {
+ if (itr->ifindex == dev->ifindex) {
+ list_del(&itr->list);
+ kfree(itr);
+ }
+ }
+
spin_unlock_bh(&dcb_lock);
}
--
2.33.0
^ permalink raw reply [flat|nested] 3+ messages in thread* Re: [PATCH net] net: dcb: flush rewrite table on device unregister
2026-11-11 3:31 [PATCH net] net: dcb: flush rewrite table on device unregister Wenbo Sun
@ 2026-11-11 8:54 ` Daniel Machon
2026-11-11 13:01 ` Petr Machata
1 sibling, 0 replies; 3+ messages in thread
From: Daniel Machon @ 2026-11-11 8:54 UTC (permalink / raw)
To: Wenbo Sun
Cc: Petr Machata, David S . Miller, Eric Dumazet, Jakub Kicinski,
Paolo Abeni, Simon Horman, Alexei Lazar, netdev, linux-kernel,
1342116488
On Wed, Nov 11, 2026 at 11:31:14AM +0800, Wenbo Sun wrote:
> In dcbnl_flush_dev(), which is called upon NETDEV_UNREGISTER for net
> devices with dcbnl_ops, only entries in dcb_app_list matching
> dev->ifindex are freed. The entries in dcb_rewr_list added via
> dcb_setrewr() are not flushed.
>
> This causes a memory leak of struct dcb_app_type nodes in dcb_rewr_list
> when a DCB-capable net device is unregistered, and leaves stale entries
> in the global rewrite table that could inappropriately match if another
> net device later reuses the same ifindex.
>
> Iterate over dcb_rewr_list in dcbnl_flush_dev() and remove and free all
> entries matching dev->ifindex, mirroring the cleanup performed for
> dcb_app_list.
>
> This issue was found via LLM-assisted code review, confirmed by
> reproducing the leaked rewrite table entry upon netdevice unregistration
> in a User-Mode Linux (ARCH=um) environment, and verified in UML by
> confirming that all rewrite table entries for the unregistered netdevice
> are flushed.
>
> Fixes: 622f1b2fae2e ("net: dcb: add new rewrite table")
> Signed-off-by: Wenbo Sun <sunwenbo@cmss.chinamobile.com>
Thanks for the fix!
I reproduced it on lan969x. After a device is unregistered, another
DCB-capable device that gets the same ifindex reports the old
device's rewrite entries in "dcb rewr show". With this patch the
table is empty.
Reviewed-by: Daniel Machon <daniel.machon@microchip.com>
Tested-by: Daniel Machon <daniel.machon@microchip.com>
^ permalink raw reply [flat|nested] 3+ messages in thread* Re: [PATCH net] net: dcb: flush rewrite table on device unregister
2026-11-11 3:31 [PATCH net] net: dcb: flush rewrite table on device unregister Wenbo Sun
2026-11-11 8:54 ` Daniel Machon
@ 2026-11-11 13:01 ` Petr Machata
1 sibling, 0 replies; 3+ messages in thread
From: Petr Machata @ 2026-11-11 13:01 UTC (permalink / raw)
To: Wenbo Sun
Cc: Daniel Machon, Petr Machata, David S . Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni, Simon Horman, Alexei Lazar, netdev,
linux-kernel, 1342116488
Wenbo Sun <sunwenbo@cmss.chinamobile.com> writes:
> In dcbnl_flush_dev(), which is called upon NETDEV_UNREGISTER for net
> devices with dcbnl_ops, only entries in dcb_app_list matching
> dev->ifindex are freed. The entries in dcb_rewr_list added via
> dcb_setrewr() are not flushed.
>
> This causes a memory leak of struct dcb_app_type nodes in dcb_rewr_list
> when a DCB-capable net device is unregistered, and leaves stale entries
> in the global rewrite table that could inappropriately match if another
> net device later reuses the same ifindex.
>
> Iterate over dcb_rewr_list in dcbnl_flush_dev() and remove and free all
> entries matching dev->ifindex, mirroring the cleanup performed for
> dcb_app_list.
>
> This issue was found via LLM-assisted code review, confirmed by
> reproducing the leaked rewrite table entry upon netdevice unregistration
> in a User-Mode Linux (ARCH=um) environment, and verified in UML by
> confirming that all rewrite table entries for the unregistered netdevice
> are flushed.
>
> Fixes: 622f1b2fae2e ("net: dcb: add new rewrite table")
> Signed-off-by: Wenbo Sun <sunwenbo@cmss.chinamobile.com>
Reviewed-by: Petr Machata <petrm@nvidia.com>
However, this has never worked, is therefore not a regression, and
should target net-next.
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2026-11-11 13:01 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2026-11-11 3:31 [PATCH net] net: dcb: flush rewrite table on device unregister Wenbo Sun
2026-11-11 8:54 ` Daniel Machon
2026-11-11 13:01 ` Petr Machata
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
all inboxes | Powered by JetHome®