On 5/28/24 12:19 PM, Sebastian Andrzej Siewior wrote: > On 2024-05-28 10:10:50 [+0200], Vlastimil Babka (SUSE) wrote: >> > I don't mind sending a patch. I'm just not sure if the lock is the right >> > thing to do. However it should ensure that interrupts are disabled on >> > !RT for the sake of the counter update (if observed in IRQ context). >> >> Looks like some places there use VM_WARN_ON_IRQS_ENABLED() that's turned off >> for PREEMPT_RT, so maybe that's what should replace the current >> lockdep_assert, perhaps together with >> lockdep_assert_held(this_cpu_ptr(&memcg_stock.stock_lock)); >> >> But also __mod_memcg_lruvec_state() already has that VM_WARN_ON. > > This "VM_WARN_ON_IRQS_ENABLED" is the initial assert for "interrupts > must be disabled while change the counter". > You want to replace it with lockdep? Part of its requirement was that it > yells with lockdep disabled. No I meant the other way around, that we'de use VM_WARN_ON_IRQS_ENABLED instead of the current lockdep_assert() that doesn't work on RT. > Currently I am leaning towards removing the > lockdep_assert_irqs_disabled() from __mod_objcg_mlstate(). Nothing but > the counter need it and they have their own check. So? Yeah that might be sufficient. > Sebastian
