<html><body>
<p><tt>dvhltc@linux.vnet.ibm.com wrote on 09/02/2010 01:04:28 AM:<br>
<br>
> Subject</tt><br>
<tt>> <br>
> Re: [PATCH][RFC] preempt_count corruption across H_CEDE call with <br>
> CONFIG_PREEMPT on pseries</tt><br>
<tt> <br>
> With this in place, we no longer see the preempt_count dropping below<br>
> zero. However, if I offline/online a CPU about 246 times I hit the<br>
> opposite problem, a preempt_count() overflow. There appears to be a<br>
> missing preempt_enable() somewhere in the offline/online paths.</tt><br>
<br>
<tt>This (preempt_count overflow) also occurred in mainline (with</tt><br>
<tt>CONFIG_PREEMPT=y) in 2.6.35, but not in 2.6.36-rc3. A </tt><br>
<tt>bisect seems to indicate it was fixed with </tt><br>
<tt>a7c2bb8279d20d853e43c34584eaf2b039de8026 "powerpc: Re-enable preemption before cpu_die()".</tt><br>
<br>
<tt>Which may look familiar. :-)</tt><br>
<br>
<tt>It looks like this patch went to mainline (likely via the powerpc tree), but may</tt><br>
<tt>have not gotten back into the -rt branch. </tt><br>
<br>
<tt>-Will</tt><br>
<tt><br>
> <br>
> -- <br>
> Darren Hart<br>
> IBM Linux Technology Center<br>
> Real-Time Linux Team<br>
</tt></body></html>