CONFIG_DEBUG_PREEMPT (was: [PATCH 04/04] Cpuset: skip rcu check ...)

[Paul Jackson <pj@sgi.com>]

pj wrote:
> instead consider removing CONFIG_DEBUG_PREEMPT from at least sn2

Ah - perhaps not so.  Adding my SGI colleague Greg Edwards to the cc
list.  My email archives suggest that he enabled CONFIG_DEBUG_PREEMPT
in ia64 sn2_defconfig a few months ago, and I presume did so intentionally.

The change enabling CONFIG_DEBUG_PREEMPT was:
    user:        Greg Edwards <edwardsg@sgi.com>
    date:        Tue Aug 16 23:38:16 2005 +0011
    summary:     [IA64] Refresh arch/ia64/configs/sn2_defconfig.

Greg - CONFIG_DEBUG_PREEMPT adds a couple of pages of assembly code 
due to various BUG checks beneath rcu_read_lock() on some hot code
paths (which is where rcu is most popular).  See the two calls
add_preempt_count() and sub_preempt_count() in kernel/sched.c.

Was this intentional to enable CONFIG_DEBUG_PREEMPT in sn2_defconfig?

Other evidence opposing this DEBUG opttion:

    Most other DEBUG options are turned off in the defconfigs.

Other evidence supporting setting this DEBUG option:

    We're not the only arch enabling CONFIG_DEBUG_PREEMPT.  See also:
        collie            simpad            s390              se7705
        lpd7a400          bigsur            dreamcast         sh03
        lpd7a404          microdev          systemh           mx1ads

-- 
                  I won't rest till it's the best ...
                  Programmer, Linux Scalability
                  Paul Jackson <pj@sgi.com> 1.925.600.0401