Re: [PATCH v3 2/2] Make hard lockup detection use timestamps

[Don Zickus <dzickus@redhat.com>]

On Wed, Aug 03, 2011 at 11:27:16AM -0700, ZAK Magnus wrote:
> An update:
> 
> I realized that the timing function I was using was no good. On x86 it
> will try to use the TSC but if it can't (which, as I understand, isn't
> a very rare case), it will fall back on a jiffies-based solution. If
> interrupts are disabled, time would appear to "stand still" and
> lockups wouldn't be detected. So, looks like I have to nix this
> particular patch and not change the hard lockup detection that much.

Hmm, I was initially worried about that, but after seeing the function
you were using assuming Peter vetted it properly to use cleanly from an
NMI context.  I guess I didn't fully understand the limitations of it.

> 
> Meanwhile, I'm working on printing stack traces to a buffer/out to
> file. I got some help and it looks like it's not that hard to do. I'll
> send it when it's ready.

Interesting.  I look forward to seeing it.

Cheers,
Don