On 2019-02-20 08:40:28 [+0100], Alain Chambron wrote: > Hello, Hi, > moreover, I have added some simple logic so that atc_tx_status does not > try to acquire the lock if it is already hold by atc_tasklet. The same > must then apply to atc_terminate_all. But once this is done, it works well. Okay. So we know that the lock is acquired recursively. Can you try the non-RT version with lockdep and post a splat? Maybe even try the latest v5.0-rc? Then it would be easier to argue in the patch that the recursive locking existing since beginning and nobody noticed it earlier because without lockdep (or RT) spin_lock() is limited to preempt_disable() and recursive locking is not noticed. > regards, > Alain Sebastian
