On Thu, 30 Apr 2015 11:40:07 -0700 Austin Schuh <austin@xxxxxxxxxxxxxxxx> wrote: > On Thu, Apr 30, 2015 at 11:32 AM, Steven Rostedt <rostedt@xxxxxxxxxxx> wrote: > > On Thu, 30 Apr 2015 20:07:21 +0200 > > Peter Zijlstra <peterz@xxxxxxxxxxxxx> wrote: > >> The irony, this is distinctly non deterministic code you're putting > >> under a RT specific preempt_disable ;-) > > > > I know :-( > > > > Unfortunately, a RT behaving fix would be much more invasive and would > > probably require the help of the xfs folks. For now, this just prevents > > a live lock that can happen and halt the system, where it becomes > > deterministic catastrophe. > > > > -- Steve > > Would it work to instead create a lock to replace the > preempt_enable_rt/preempt_disable_rt pair in XFS? Not just the place where the preempt_enable and disable is done, but it would need to replace all the per cpu bit spin locks (the XFS_ICSB_FLAG_LOCK bit in icsbp->icsb_flags). If we can replace them with a rtmutex (spin_lock() in vanilla kernel), then that would work. -- Steve -- To unsubscribe from this list: send the line "unsubscribe linux-rt-users" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
