This is a discussion on Re: getaffinity/setaffinity and cpu sets. - FreeBSD ; On Thu, 21 Feb 2008, David Schultz wrote: > I have no specific comments, but I wanted to point out that the > Solaris kernel team put a lot of thought into coming up with a > flexible processor binding ...
On Thu, 21 Feb 2008, David Schultz wrote:
> I have no specific comments, but I wanted to point out that the
> Solaris kernel team put a lot of thought into coming up with a
> flexible processor binding API for Solaris 10 that meshes well
> with jails (a.k.a. zones in Solaris). It might be worthwhile to
> investigate what good ideas they might have had, and to decide if
> compatibility is worthwhile:
Well interestingly enough I came up with an almost identical thing but
with fewer features. If someone wanted to add the missing features I'm
fine with that. I don't know if we want to or are free to copy the api
> There are manpages for this stuff somewhere. pset_create and
> pset_destroy obviously create and delete processor set definitions.
> pset_bind binds a particular process / thread / session / jail
> to a processor set, similar to your CPU_WHICH_* flags, I think,
> but with more options.
I didn't see the header but I read some web articles about it. The only
real discrepancy is whether you explicitly destroy them or they disappear
when the last processor leaves the set. I could be convinced to do
either, but so far I'm leaning towards automatic destruction.
Hopefully someone who is involved with jails will work with me to
integrate the two. I intend only to implement the infrastructure required
to create and modify sets. Someone who is jail savvy will need to apply
the set and the appropriate security restrictions.
email@example.com mailing list
To unsubscribe, send any mail to "firstname.lastname@example.org"