Re: 2.6.22-rc2-mm1

[Ingo Molnar <mingo@elte.hu>]

* Christoph Lameter <clameter@sgi.com> wrote:

> +SLUB Debug output
> +-----------------
> +
> +Here is a sample of slub debug output:
> +
> +*** SLUB kmalloc-8: Redzone Active@0xc90f6d20 slab 0xc528c530 offset=3360 flags=0x400000c3 inuse=61 freelist=0xc90f6d58
> +  Bytes b4 0xc90f6d10:  00 00 00 00 00 00 00 00 5a 5a 5a 5a 5a 5a 5a 5a ........ZZZZZZZZ
> +    Object 0xc90f6d20:  31 30 31 39 2e 30 30 35                         1019.005
> +   Redzone 0xc90f6d28:  00 cc cc cc                                     .
> +FreePointer 0xc90f6d2c -> 0xc90f6d58
> +Last alloc: get_modalias+0x61/0xf5 jiffies_ago=53 cpu=1 pid=554
> +Filler 0xc90f6d50:  5a 5a 5a 5a 5a 5a 5a 5a                         ZZZZZZZZ

when constructing debug output it's a good idea to make it as 
'context-free' as possible. I.e. instead of saying 'Filler' and 
explaining it somewhere in Documentation/*, just say:

  Expected memory values at 0xc90f6d50:  5a 5a 5a 5a 5a 5a 5a 5a
  ... but found the following values:    12 34 56 78 90 12 34 56

(or whatever it means)

please also put in one of the keywords as the first word of whatever 
SLUB prints:

   BUG: ...
   WARNING: ...
   INFO: ...

the output should be compact, but never at the price of 
understandability. For example i tried to make lockdep output as 
self-explanatory as possible - and a good portion of lockdep.c is about 
having around the kind of information to make the error more 
understandable (tracking lock-acquire stacktraces, tracking 
initialization points, tracking initialization-time symbol names, etc.), 
and there's even quite some algorithmic complexity from that. I probably 
spent more time on making lockdep output easier to decode than on any 
other sub-feature of lockdep. The output format literally went over 
hundreds of revisions. Output to humans is _the_ most important thing 
that a debug facility can do, sometimes more important than the 
debugging it offers ;-) And even the existing lockdep output sometimes 
confuses people.

	Ingo