[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-devel] [PATCH] x86/watchdog: Use real timestamps for watchdog timeout

At 14:17 +0100 on 24 May (1369405052), Jan Beulich wrote:
> > Using TSC to gate the actual watchdog crash might get a bit messy,
> > especially if it ends up adding code to the users of write_tsc().
> The only problematic write_tsc() user is that to recover from a
> stopped counter in deep C states. This is not a meaningful problem
> because - as just said in another reply - NMI storms and long times
> spent in deep C states exclude each other.

Yep.  But we could go into a C-state and be woken by an NMI storm,
meaning that our 'before' TSC value would be immediately wrong.  The
more I think about this the less I think we should be trying to measure
elapsed time except to change the printout.

> > And nmi_watchdog_tick() can just check regs->eip as a
> > hint not to trust the scale factors. :)
> By doing a range check looking for it to point into some function?
> How would that cope with LTO?

Unreliably. :)  I imagine in most cases it would be fine, since that
function has its address taken, but unless we called out to an explicit
.S routine to do the update it ouwld be at best a hint. 


Xen-devel mailing list



Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.