Subject: Re: splraiseipl()
To: YAMAMOTO Takashi <email@example.com>
From: Jason Thorpe <firstname.lastname@example.org>
Date: 11/01/2005 09:11:32
On Nov 1, 2005, at 5:39 AM, YAMAMOTO Takashi wrote:
> splraiseipl() is an MD/MI interface found in newlock branch.
> basically splraiseipl(IPL_BIO) == splbio().
> i'd like to see it in trunk, because i'd like to use it for workqueue.
> (and maybe for callout and kcont.)
I'd like to see it in the trunk, too! :-)
> now, some questions before digging much into MD code:
> - it (ab)uses IPL_ values which is used for eg. pci_intr_establish().
> is it worth to introduce another set of definitions eg. SPL_BIO?
IPL_* are the appropriate values to use, I think. IPL_xxx should
correspond directly to splxxx(), so there is no need for additional
constants that would basically be aliases of IPL_*.
> - a developer said its name ("splraiseipl") is weired.
> any better name suggestions?
Maybe just "spl(xxx)"?
> - which set of splxxx is considered be usable from MI code?
> eg. while spl(9) mentions about splserial, it doesn't seem to be
> available for some ports.
I think IPL_SERIAL should be an MI definition. If a given platform
doesn't support it, I suppose we could always use #ifdef.