[lkml]   [2015]   [Feb]   [19]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
SubjectRe: [PATCH 32/35] clockevents: Fix cpu down race for hrtimer based broadcasting
On 02/18/2015 06:36 PM, Peter Zijlstra wrote:
> On Wed, Feb 18, 2015 at 08:40:52AM +0530, Preeti U Murthy wrote:
>> Look at the changelog,
> Heh, yah, clearly I tl;dr'd that. Indeed.
>> it explains why tick_takeover must be called
>> *before* __cpu_die().
> Since you reported this, can you test if things work if you move that
> function call to before __cpu_die() ?

Yes it does. I had tested and posted out a patch to fix the issue
here: I see now that tglx proposes
to get rid of "taking over" tick duties going forward, rather "handing
over" tick duties would be better. So the patch in the above link would
not be acceptable. But minus the taking over of the do_timer, the cpu
doing the hotplug takes over broadcast duty before __cpu_die() in this
patch. This design works.

Preeti U Murthy
> I _think_ it should be fine, but what do I know ;-)

 \ /
  Last update: 2015-02-19 08:21    [W:0.127 / U:0.068 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site