5.10-stable review patch. If anyone has any objections, please let me know. ------------------ From: Michal Luczaj <mhal@xxxxxxx> commit 75e072a390da9a22e7ae4a4e8434dfca5da499fb upstream. Consider a sockmap entry being updated with the same socket: osk = stab->sks[idx]; sock_map_add_link(psock, link, map, &stab->sks[idx]); stab->sks[idx] = sk; if (osk) sock_map_unref(osk, &stab->sks[idx]); Due to sock_map_unref(), which invokes sock_map_del_link(), all the psock's links for stab->sks[idx] are torn: list_for_each_entry_safe(link, tmp, &psock->link, list) { if (link->link_raw == link_raw) { ... list_del(&link->list); sk_psock_free_link(link); } } And that includes the new link sock_map_add_link() added just before the unref. This results in a sockmap holding a socket, but without the respective link. This in turn means that close(sock) won't trigger the cleanup, i.e. a closed socket will not be automatically removed from the sockmap. Stop tearing the links when a matching link_raw is found. Fixes: 604326b41a6f ("bpf, sockmap: convert to generic sk_msg interface") Signed-off-by: Michal Luczaj <mhal@xxxxxxx> Signed-off-by: Daniel Borkmann <daniel@xxxxxxxxxxxxx> Reviewed-by: John Fastabend <john.fastabend@xxxxxxxxx> Link: https://lore.kernel.org/bpf/20241202-sockmap-replace-v1-1-1e88579e7bd5@xxxxxxx Signed-off-by: Greg Kroah-Hartman <gregkh@xxxxxxxxxxxxxxxxxxx> --- net/core/sock_map.c | 1 + 1 file changed, 1 insertion(+) --- a/net/core/sock_map.c +++ b/net/core/sock_map.c @@ -161,6 +161,7 @@ static void sock_map_del_link(struct soc verdict_stop = true; list_del(&link->list); sk_psock_free_link(link); + break; } } spin_unlock_bh(&psock->link_lock);