On Thu, May 17, 2007 at 10:02:43AM +0200, giovanni wrote: > I was very curious about the implementation so I've read it. > http://www.openbsd.org/cgi-bin/cvsweb/src/sys/arch/i386/i386/mp_setperf.c?rev=1.2&content-type=text/x-cvsweb-markup
> well I was wondering why is it necessary the splipi when before we are > at IPL_HIGH? The code was orignally written without the mutex using the MD i386_atomic_testset_i() because I was worried (eerroneously) about highly contrived corner case that might result in deadlock. After talking to art it was clear I was on crack and a mutex would do fine. This code is likely vestigial and can be removed I will look into it. gwk > ... > s = splipi(); > ul_setperf(mp_setperflevel); > splx(s); > ... > tkanks,
