Re: multiple vcpus in a domain

[John Nemeth <jnemeth%cue.bc.ca@localhost>]

On Mar 11, 10:03, Manuel Bouyer wrote:
} On Wed, Mar 10, 2021 at 08:11:25PM -0800, John Nemeth wrote:
} > On Mar 10, 13:56, Greg Troxel wrote:
} > } 
} > } I wrote in the HOWTO that as of current NetBSD dom0 supports multiple
} > } CPUs, and want to check that this is right.
} > } 
} > } I think:
} > } 
} > }    8, 9, current XEN3_DOMU all can have multiple vcpus and use them and
} > }    this is basically working fine
} > 
} >      domU has been multiprocessor since 6.0.
} > 
} > }    8, 9 XEN3_DOM0 are not built with options MULTIPROCESSOR and if given
} > }    >1 vcpu will just ignore all but one
} > 
} >      This is likely not true.  I ran a test server with dom0
} > multiprocessor some time ago.
} 
} You've been lukcy then. Some parts of dom0 support is not MP-safe in -9.

     Who said anything about -9?  :->  It would have been -8 or
even earlier when cherry@ first asked for testers.

i386devel: {527} what /netbsd.xen0gen.smp
/netbsd.xen0gen.smp
        NetBSD 7.99.32 (XEN3_DOM0) #0: Thu Jul 14 16:20:31 PDT 2016

Note: old hostname, the box has been amd64 for a long time.

} > }    current XEN3_DOM0 is built with MULTIPROCESSOR, and detects and uses
} > }    multiple vcpus and everything is fine
} > 
} >      It will use multiple cpus, but I'm not certain that everything
} > is fine.  Last I knew, there was still some work to be done on
} > this, but I may be out of date here.  I would be very happy to be
} > wrong here, since single processor dom0 is likely to be a major
} > bottleneck.
} 
} dom0 is supposed to be MP-safe in HEAD now.

     This is good news.  I won't be running it on production servers
any time soon, but hopefully it pans out.

}-- End of excerpt from Manuel Bouyer