On Fri, May 26, 2017 at 10:31:03AM +1000, Nicholas Piggin wrote: > On Thu, 25 May 2017 09:55:59 -0400 > Don Zickus <dzickus@xxxxxxxxxx> wrote: > > > On Thu, May 25, 2017 at 06:28:54PM +1000, Nicholas Piggin wrote: > > > For architectures that define HAVE_NMI_WATCHDOG, instead of having > > > them provide the complete touch_nmi_watchdog() function, just have > > > them provide arch_touch_nmi_watchdog(). > > > > > > This gives the generic code more flexibility in implementing this > > > function, and arch implementations don't miss out on touching the > > > softlockup watchdog or other generic details. > > > > The idea makes sense. I don't think you can have hld_touch_nmi_watchdog > > defined with arch_touch_nmi_watchdog, so I am wondering if it makes sense to > > combine them somehow. Though renaming hld_touch_nmi_watchdog to > > arch_touch_nmi_watchdog sounds odd, I think it mimics the idea. > > Yeah I agree it's not quite right, and I think using > arch_touch_nmi_watchdog would be fine for the hld, which makes sense > if you think of it as a utility or library function for architectures > that want a hardlockup watchdog and can use perf for it. Yeah, if you wouldn't mind trying that. Over the last year it seems there is a push to make the hld more of a separate thing if folks want to use perf. I have been trying to tweak it so it can be used in-place of the arch solution or just use the arch solution. And still retain the same function calls. Cheers, Don > > I can change that if you prefer. BTW the 0day picked up another > Kconfig compile bug, so I'll respin the series and include any changes > you like. > > Thanks, > Nick
