On Mon, Oct 25, 2021 at 12:44:05PM +0200, Vlastimil Babka wrote: > On 10/19/21 14:50, Matthew Wilcox (Oracle) wrote: > > Leave interrupts disabled when we change which lru lock is held. > > > > Signed-off-by: Matthew Wilcox (Oracle) <willy@xxxxxxxxxxxxx> > > Acked-by: Vlastimil Babka <vbabka@xxxxxxx> > > Assuming lockdep is fine with e.g.: > > spin_lock_irqsave(A); > spin_unlock(A); > spin_lock(B); > spin_unlock_irqrestore(B); > > (with A and B of same class). It's unconditionally okay with that pattern. As far as lockdep is concerned there really is no differences vs: local_irq_save() spin_lock(a) spin_unlock(a) spin_lock(b) spin_unlock(b) local_irq_restore() It's the RT locking primitives that care about the difference :-)
