cvs commit: src/sys/netgraph ng_base.c src/sys/sys kernel.h

[Scott Long]

Simon L. Nielsen wrote:

> On 2005.01.10 13:28:17 -0700, Scott Long wrote:
> 
>>Simon L. Nielsen wrote:
>>
>>
>>>On 2005.01.10 10:41:39 -0800, John-Mark Gurney wrote:
>>>
>>>
>>>>So, to everone, MODULE_DEPEND only marks which modules are required
>>>>for this module to work, it does not infulence startup ordering at
>>>>all...
>>>
>>>So, should I commit this patch?
>>>
> 
> [...]
> 
>>>@@ -69,3 +69,7 @@
>>>.An -nosplit
>>>This manual page was written by
>>>.An Alexander Langer Aq alex at FreeBSD.org .
>>>+.Sh BUGS
>>>+.Nm
>>>+only marks which modules are required for the module to work, it does
>>>+not influence start-up ordering.
>>
>>It also is used by the loader and by kldload to load the dependent 
>>modules.  So it's more than just an informational tag, it actually does
>>get used by the system.
> 
> 
> I don't think the above says it's only informational, but I'm sure the
> wording could be improved.  Perhpaps somebody has a suggestion for
> better describing the issue?
> 

Well, I guess my point is that it's important to document exactly what
MODULE_DEPEND does do, and well as what it doesn't do.  It's not really
a bug that it doesn't change SI_ORDER order, it's just a not a feature
that it was ever intended to support.  A second paragraph under 
DESCRIPTION that talks about this would probably be most appropriate.
Something like:

.Nm
provides hints to the kernel
.Xr loader 8
and to
.Xr kldload 8
to ensure that the named dependency is loaded prior to the existing 
module.  It does not change or dictate the order in which modules are
initialized at runtime.

Scott