Re: mmotm 2009-04-10-02-21 uploaded - forkbombed by work_for_cpu

[Andrew Morton <akpm@linux-foundation.org>]

On Mon, 13 Apr 2009 16:50:45 -0700 (PDT) Linus Torvalds <torvalds@linux-foundation.org> wrote:

> 
> 
> So I applied this (commit 01599fca6758d2cd133e78f87426fc851c9ea725: 
> "cpufreq: use smp_call_function_[single|many]() in acpi-cpufreq.c"), but 
> just realized - because of a compiler warning - that this looks 
> suspicious:
> 
> On Mon, 13 Apr 2009, Andrew Morton wrote:
> > @@ -283,7 +280,7 @@ static unsigned int get_measured_perf(st
> >  	unsigned int perf_percent;
> >  	unsigned int retval;
> >  
> > -	if (!work_on_cpu(cpu, read_measured_perf_ctrs, &readin))
> > +	if (smp_call_function_single(cpu, read_measured_perf_ctrs, &cur, 1))
> >  		return 0;
> >  
> >  	cur.aperf.whole = readin.aperf.whole -
> 
> How and why did that "read_measured_perf_ctrs, &readin" become 
> "read_measured_perf_ctrs, &cur" when the work_on_cpu() was converted to 
> "smp_call_function_single()"?
> 
> Looks like a bug. But such an odd one that I wonder whether there was some 
> thought behind it? Andrew? 
> 

<scratches head>

OK, the acpi tree went and had conflicting changes merged into it after
I'd written the patch:

@@ -281,52 +279,57 @@ static long read_measured_perf_ctrs(void
 static unsigned int get_measured_perf(struct cpufreq_policy *policy,
 				      unsigned int cpu)
 {
-	struct perf_cur cur;
+	struct perf_pair readin, cur;
 	unsigned int perf_percent;
 	unsigned int retval;
 
-	if (!work_on_cpu(cpu, read_measured_perf_ctrs, &cur))
+	if (!work_on_cpu(cpu, read_measured_perf_ctrs, &readin))
 		return 0;
 
+	cur.aperf.whole = readin.aperf.whole -
+				per_cpu(drv_data, cpu)->saved_aperf;
+	cur.mperf.whole = readin.mperf.whole -
+				per_cpu(drv_data, cpu)->saved_mperf;
+	per_cpu(drv_data, cpu)->saved_aperf = readin.aperf.whole;
+	per_cpu(drv_data, cpu)->saved_mperf = readin.mperf.whole;
+

and it appears that I incorrectly reverted part of
18b2646fe3babeb40b34a0c1751e0bf5adfdc64c while fixing the resulting
rejects.

Switching it to `readin' looks correct.