Subject: splhigh() is NOT ALWAYS bad
To: None <tech-kern@NetBSD.ORG>
From: Gordon W. Ross <email@example.com>
Date: 01/21/1998 10:37:14
> Date: 21 Jan 1998 06:00:02 -0000
> From: source@NetBSD.ORG
> Tue Jan 20 21:54:40 PST 1998
> Update of /cvsroot/src/sys/arch/sparc/dev
> In directory netbsd1:/var/slash-tmp/cvs-serv7694
> Modified Files:
> Log Message:
> KNF, and use splzs over splhigh in one place.
I don't mean to pick on mrg specifically here, because I'm
sure the change was well-meant, but I'd like to point out
that the use of splhigh is _not_always_ bad, expecially
when that state is held for a very short time. Further
using splhigh is convenient in code that must be called
at different interrupt protections, because there is no
risk of accidentally lowering the interrupt level.
With that explained, it should be clear that I intentionally
used splhigh in zs_attach to avoid lowering the interrupt
level during the autoconfig phase.
In summary, let's avoid "splhigh() is always bad" religion.