On 2018-03-23 16:25:25 [+0000], Bart Van Assche wrote: > On Fri, 2018-03-23 at 16:55 +0100, Sebastian Andrzej Siewior wrote: > > I am going take this into -RT tree for now until we have different > > solution. > > Have you considered to delete the WARN_ON_ONCE(!irqs_disabled()) statement? > I think that check duplicates functionality that already exists in lockdep > since lockdep is already able to detect spinlock use inconsistencies. correct. That is why I suggested to use lockdep_assert_held() instead of this IRQ-check + the spin_lock_assert(). The only downside is that this code (as of now) works with lockdep disabled. On the other hand, lockdep_assert_held() gives you a splat instead of a BUG() statement like spin_lock_assert() does so I clearly promote lockdep here :) > Bart. Sebastian
