Re: [Patch 1/2] Trace code and documentation (updated)

[David Wilder <dwilder@us.ibm.com>]

Andi Kleen wrote:
> "David J. Wilder" <dwilder@us.ibm.com> writes:
> 
> Not having read the whole thing; just something I noticed.
> 
> Gut feeling is that you have too many knobs and options and 
> some overengineering though -- simplifying it would be a good thing.
> 
>> +
>> +#define TRACE_PRINTF_TMPBUF_SIZE (1024)
>> +static char trace_tmpbuf[NR_CPUS][TRACE_PRINTF_TMPBUF_SIZE];
> 
> That definitely needs to be a per CPU variable. Imagine
> what happens on a NR_CPUS==4096 kernel.  In general when 
> you have a NR_CPUS indexed array you're likely doing something
> wrong. Yes there are still places in the main tree who do that,
> but most of them need to be fixed.

I agree with you; however, this is in the example code in the 
Documentation directory,  It is not part of the trace code.  The example 
was just meant to be a demonstration of how the interface works.
> 
> -Andi
>