Hello, I think, the problem should be in step 2 below: - step 1: try_stop_module -> try_release_module_ref -> atomic_sub_return sets module refcnt to 0. After this point the process to remove module
hello Julian, thanks a lot for reviewing this code. [...] Initially I believed that __ip_vs_dev_cleanup() was called when start_sync_thread() was already running (but it already released the RTNL loc
Hello, May be this is not early enough. When control hooks are the first thing to disconnect with ip_vs_unregister_nl_ioctl() how start_sync_thread() is called concurrently? ip_vs_sockopts is protect
if the IPVS module is removed while the sync daemon is starting, there is a small possibility for stop_sync_thread() to end successfully just after start_sync_thread() released sync_mutex, but before