config(8) should check if a scheduler is selected
Wes Peters
wes at softweyr.com
Sun Apr 20 16:53:08 PDT 2003
On Saturday 19 April 2003 01:05, Scott Long wrote:
> Bruce Evans wrote:
> >
> > It is the only mandatory option (sic). Kernels with no options
> > (although they might not be useful) can be built except for this bug.
> > Example of a minimal config file (before misconfiguration of the
> > configuration of scheduling).
> >
> > %%%
> > machine i386
> > cpu I686_CPU
> > ident MIN
> > %%%
>
> The scheduler is (one of) the first core subsystems to be made
> modular. If by chance the VM system became modular (VM_MACH, VM_UVM
> =-) you'd have a similar situation there also.
Doesn't this argue for a keyword rather than an option? If you have to
have one or the other for the kernel to function, wouldn't a 'scheduler'
keyword (and likewise a 'vm' or 'vm_model' keyword) save us from the
lunacy of non-optional options?
> I'm afraid that the lack of seatbelts in config(8) for SCHED_xxx will
> generate a lot of user complaints when 5.1 is released. Since code to
> implement it has not magically appeared yet, we might have to make due
> with adding extra eye-catching comments to things like NOTES and
> GENERIC.
Or maybe we could fix it?
> > BTW, a minimal kernel is now almost 3 times as large as in FreeBSD-2
> > due to general bloat and misconfiguration of configuration in the
> > opposite way (subsystems much larger than scheduling are standard;
> > you can still leave out FFS and INET but many less useful subsystems
> > are standard).
>
> Some of us remember when 250k FreeBSD kernels were not hard to
> configure =-)
And 330K kernels were the norm, as long as you eschewed NFS. Sigh.
--
Where am I, and what am I doing in this handbasket?
Wes Peters wes at softweyr.com
More information about the freebsd-arch
mailing list