[Virtio-fs] virtio-fs on zLinux (s390x)

Dr. David Alan Gilbert dgilbert at redhat.com
Tue Mar 17 12:11:47 UTC 2020


* Igor Mammedov (imammedo at redhat.com) wrote:
> On Tue, 17 Mar 2020 02:05:06 -0500
> "Gong Su" <gongsu at us.ibm.com> wrote:
> 
> > 
> > 
> > Hi Igor, just wondering if there is any updates on getting the patches that
> > will allow shared memory without -numa (not supported on s390 yet)? Thanks.
> 
> it's already upstream.
> 
> there is a new upstream option -machine memory-backend=...
> that allows to set custom backend for main RAM

SOmething like:
   -object memory-backend-memfd,size=2G,id=mem,share=on
   -machine pc,memory-backend=mem

seems to be the syntax then.

Good, that does make life simpler.

Dave

> > __________________
> > 
> > Hi Igor, that would be great. Thanks.
> > 
> > 
> > 
> > From:	Igor Mammedov <imammedo at redhat.com>
> > To:	"Gong Su" <gongsu at us.ibm.com>
> > Cc:	"Dr. David Alan Gilbert" <dgilbert at redhat.com>, Cornelia Huck
> >             <cohuck at redhat.com>, virtio-fs at redhat.com
> > Date:	12/13/2019 08:17 AM
> > Subject:	[EXTERNAL] Re: [Virtio-fs] virtio-fs on zLinux (s390x)
> > 
> > 
> > 
> > On Thu, 12 Dec 2019 14:10:32 -0500
> > "Gong Su" <gongsu at us.ibm.com> wrote:
> > 
> > >
> > > Thanks Dave. That was quick!
> > >
> > > Igor, any idea when this will happen? :-)
> > If you need early sneak peek, I can post common + s390 patches.
> > 
> > 
> > > From:		 "Dr. David Alan Gilbert" <dgilbert at redhat.com>
> > > To:		 Gong Su <gongsu at us.ibm.com>
> > > Cc:		 virtio-fs at redhat.com, Cornelia Huck <cohuck at redhat.com>
> > > Date:		 12/12/2019 01:59 PM
> > > Subject:		 [EXTERNAL] Re: [Virtio-fs] virtio-fs on zLinux (s390x)
> > >
> > >
> > >
> > > * Gong Su (gongsu at us.ibm.com) wrote:
> > > >
> > > >
> > > > Hi Stefan, I'm resending my questions to the list per your request.
> > > >
> > > > Basically, I'm trying to see if I can get virtio-fs to work on zLinux.
> > > > Following the instructions on
> > >
> > https://urldefense.proofpoint.com/v2/url?u=https-3A__virtio-2Dfs.gitlab.io_howto-2Dqemu.html&d=DwIFAg&c=jf_iaSHvJObTbx-siA1ZOg&r=tf-1fDxgEPUcSzS0GcREznaRLLY6eGivTqJEnr2bEmQ&m=bKfrjedGnP-cXu2BPH6ynhTbdi5nm0VW1nx6Qb5Q_7M&s=pkvbWd3P-tWRpKPww5iwoPHLE_HnkenEEJhywABpRuc&e=
> > 
> > >   I
> > > > was able to compile all the pieces but when I tried to run qemu I
> > > realized
> > > > that it needs -numa option which isn't supported on s390x yet.
> > > >
> > > > I have talked to Christian Borntraeger (whom I worked with on other
> > > > projects before) and he indicated that he had a plan to add -numa
> > support
> > > > for s390x but currently is busy on other things.
> > > >
> > > > So I'm trying to see if there are other ways for me to go forward
> > without
> > > > having to wait for Christian. I see two options:
> > > >
> > > > 1. Merge virtiofsd into qemu itself. In the "Interchangeability"
> > section
> > > of
> > > > the virtio-fs design page (
> > >
> > https://urldefense.proofpoint.com/v2/url?u=https-3A__virtio-2Dfs.gitlab.io_design.html&d=DwIFAg&c=jf_iaSHvJObTbx-siA1ZOg&r=tf-1fDxgEPUcSzS0GcREznaRLLY6eGivTqJEnr2bEmQ&m=bKfrjedGnP-cXu2BPH6ynhTbdi5nm0VW1nx6Qb5Q_7M&s=2VrG3dsgZcdW8uO6ZBiGtRgNp4X1z3d8TMXnUdHhfCc&e=
> > 
> > >  ), it
> > > > mentioned that "It is also possible to forego vhost-user and emulate
> > the
> > > > virtio-fs device directly inside the hypervisor, although this may
> > result
> > > > in poor isolation and security." Since each guest typically gets its
> > own
> > > > qemu process, maybe this is acceptable for some use cases. In exchange,
> > > you
> > > > get the benefit of not having to carve out a chunk of memory
> > > from /dev/shm
> > > > (and maybe simpler communication between virtiofsd and qemu).
> > > > 2. Use something other than -numa to share memory between virtiofsd and
> > > > qemu that's supported on zLinux. I'm not entirely familiar with qemu so
> > I
> > > > have no idea whether this is possible or not.
> > >
> > > Hi,
> > >   My understanding is that Igor Mammedov (imammedo at redhat.com) has plans
> > > to allow 'shared' to be set on main memory, so we wouldn't need the numa
> > > code.
> > >
> > > Dave
> > >
> > > > Thanks.
> > >
> > > > _______________________________________________
> > > > Virtio-fs mailing list
> > > > Virtio-fs at redhat.com
> > > >
> > >
> > https://urldefense.proofpoint.com/v2/url?u=https-3A__www.redhat.com_mailman_listinfo_virtio-2Dfs&d=DwIFAg&c=jf_iaSHvJObTbx-siA1ZOg&r=tf-1fDxgEPUcSzS0GcREznaRLLY6eGivTqJEnr2bEmQ&m=bKfrjedGnP-cXu2BPH6ynhTbdi5nm0VW1nx6Qb5Q_7M&s=O41fLQGJCOR0-Aj2Y4nziy3aTYTOCzaH8Lahe7xhBK8&e=
> > 
> > >
> > >
> > > --
> > > Dr. David Alan Gilbert / dgilbert at redhat.com / Manchester, UK
> > >
> > >
> > >
> > >
> > 
> > 
> > 
> 
--
Dr. David Alan Gilbert / dgilbert at redhat.com / Manchester, UK




More information about the Virtio-fs mailing list