Re: [PATCH 2/13] ATA ACPI: debugging infrastructure

[Andrew Morton <akpm@osdl.org>]

James Courtier-Dutton <James@superbug.co.uk> wrote:
>
> Is there a particular debugging coding style that we should adopt for
>  all the kernel code.

Err, probably.  But we'd need to have a 1000-email argument first.

Right now many subsystems and often many individual drivers go and
implement their own set of debugging macros and knobs to twiddle.  This was
a great source of fun for me in trying to support gcc-2.95.x - each time a
new debug macro got implemented I had to go in there (again) and apply the
gcc-2.95.x-macro-expansion-bug-workaround to it.

Yes, one common toolset with a common way of controlling it would be much
more sensible than the present chaos.  I count 163 separate definitions of
dprintk(), and that's excluding all the non-x86 arch and include dirs.

>  For example,
>  kconfig option in order to compile a module/section of core code for
>  debug work.
>  A sysfs file to then control the debug level for each module.
>  A debug module option, in the cases where a particular level of debug is
>  required at module load time, and before the sysfs entry exists.
>  If particularly fine grained debug control is needed, the module could
>  have multiple entries in the sysfs to control different classes of debug
>  output.
> 

Something like that..   Just don't cc me while you work it out ;)