Messages in this thread | | | Date | Thu, 5 Feb 2015 08:58:20 -0800 | From | "Paul E. McKenney" <> | Subject | Re: linux-next: Tree for Feb 4 |
| |
On Thu, Feb 05, 2015 at 03:57:12PM +0100, Sedat Dilek wrote: > On Thu, Feb 5, 2015 at 8:14 AM, Dave Hansen <dave@sr71.net> wrote: > > On 02/04/2015 05:53 PM, Sedat Dilek wrote: > >> The architecture-specific switch_mm() function can be called by offline > >> CPUs, but includes event tracing, which cannot be legally carried out > >> on offline CPUs. This results in a lockdep-RCU splat. This commit fixes > >> this splat by omitting the tracing when the CPU is offline. > > ... > >>>> >> > load_cr3(next->pgd); > >>>> >> > - trace_tlb_flush(TLB_FLUSH_ON_TASK_SWITCH, TLB_FLUSH_ALL); > >>>> >> > + if (cpu_online(smp_processor_id())) > >>>> >> > + trace_tlb_flush(TLB_FLUSH_ON_TASK_SWITCH, TLB_FLUSH_ALL); > > > > Is this, perhaps, something that we should be doing in the generic trace > > code so that all of the trace users don't have to worry about it? Also, > > this patch will add overhead to the code when tracing is off. It would > > be best if we could manage to make the cpu_online() check only in the > > cases where the tracepoint is on. > > Hi Dave, > > thanks for your feedback. > > I have just seen that I again see the call-trace.
When you get well, could you please send that call trace?
> Maybe you can discuss with Paul and others or offer a proposal patch.
The other possibility is to have a CONFIG_ARCH_DYING_IDLE or some such that allows this particular flavor of x86 to invoke the CPU_DYING_IDLE from after the call to switch_mm(). Dave, does that make sense?
My guess would be that there should be a cpu_dying_idle_generic() invoked from cpu_idle_loop(), and a cpu_dying_idle_native() invoked at the end of idle_task_exit(). Or can I get away with just moving the current rcu_notify_cpu() call from cpu_idle_loop() to the end of idle_task_exit()?
A quick look at the calls to idle_task_exit() makes this look plausible. There are a number of calls to printk() and to complete() that need help, but that is a pre-existing issue in any case, as both these code paths have RCU readers that are having no effect on offline CPUs.
Dave, thoughts?
> I should really do something for my recovery (influenza). > Instead of laying lazy in my bed I thought to update my Linux kernels > and graphics driver stack which made me happy.
Get well, being sick is bad for your health! ;-)
Thanx, Paul
> Regards, > - Sedat - >
| |