Re: [Qemu-ppc] [PATCH 8/9] spapr: Limit available pagesizes to provide a consistent guest environment

[David Gibson]

On Thu, Jun 21, 2018 at 09:01:27AM +0200, Cédric Le Goater wrote:
> On 06/18/2018 08:36 AM, David Gibson wrote:
> > KVM HV has some limitations (deriving from the hardware) that mean not all
> > host-cpu supported pagesizes may be usable in the guest.  At present this
> > means that KVM guests and TCG guests may see different available page sizes
> > even if they notionally have the same vcpu model.  This is confusing and
> > also prevents migration between TCG and KVM.
> > 
> > This patch makes the environment consistent by always allowing the same set
> > of pagesizes.  Since we can't remove the KVM limitations, we do this by
> > always applying the same limitations it has, even to TCG guests.
> > 
> > Signed-off-by: David Gibson <address@hidden>
> >
> > ---
> >  hw/ppc/spapr_caps.c | 33 +++++++++++++++++++++++++++++++++
> >  1 file changed, 33 insertions(+)
> > 
> > diff --git a/hw/ppc/spapr_caps.c b/hw/ppc/spapr_caps.c
> > index 9fc739b3f5..0584c7c6ab 100644
> > --- a/hw/ppc/spapr_caps.c
> > +++ b/hw/ppc/spapr_caps.c
> > @@ -334,6 +334,38 @@ static void 
> > cap_hpt_maxpagesize_apply(sPAPRMachineState *spapr,
> >      spapr_check_pagesize(spapr, qemu_getrampagesize(), errp);
> >  }
> >  
> > +static bool spapr_pagesize_cb(void *opaque, uint32_t seg_pshift, uint32_t 
> > pshift)
> > +{
> > +    unsigned maxshift = *((unsigned *)opaque);
> > +
> > +    assert(pshift >= seg_pshift);
> 
> you could check that elsewhere.

Um.. I'm not sure what you're getting at.

> > +    /* Don't allow the guest to use pages bigger than the configured
> > +     * maximum size */
> > +    if (pshift > maxshift) {
> > +        return false;
> > +    }
> > +
> > +    /* For whatever reason, KVM doesn't allow multiple pagesizes
> > +     * within a segment, *except* for the case of 16M pages in a 4k or
> > +     * 64k segment.  Always exclude other cases, so that TCG and KVM
> > +     * guests see a consistent environment */
> > +    if ((pshift != seg_pshift) && (pshift != 24)) {
> > +        return false;
> > +    }

Note the stanza above, I'll refer to it below.

> > +
> > +    return true;
> > +}
> 
> So, do we really need ppc_hash64_filter_pagesizes() to have a callback ? 

I agree that it seems overly involved, but it was the best way I could
see to logically separate the TCG / softmmu specific logic from the
spapr specific logic.

> It seems that we only use the routine once in the patchset and that the
> only thing we need to check is 'maxshift'.

Not quite.  An earlier draft had this routine just take a max page
size and clamp accordingly.  But that failed when I wrote the code to
check against the KVM capabilities, because KVM also excludes some
other pagesize combinations.  That's what the stanza I point out above
is about

> Do you envision other usage of the routine ?

Not really, no.

> 
> Thanks,
> 
> C.
> 
> > +static void cap_hpt_maxpagesize_cpu_apply(sPAPRMachineState *spapr,
> > +                                          PowerPCCPU *cpu,
> > +                                          uint8_t val, Error **errp)
> > +{
> > +    unsigned maxshift = val;
> > +
> > +    ppc_hash64_filter_pagesizes(cpu, spapr_pagesize_cb, &maxshift);
> > +}
> > +
> >  sPAPRCapabilityInfo capability_table[SPAPR_CAP_NUM] = {
> >      [SPAPR_CAP_HTM] = {
> >          .name = "htm",
> > @@ -401,6 +433,7 @@ sPAPRCapabilityInfo capability_table[SPAPR_CAP_NUM] = {
> >          .set = spapr_cap_set_pagesize,
> >          .type = "int",
> >          .apply = cap_hpt_maxpagesize_apply,
> > +        .cpu_apply = cap_hpt_maxpagesize_cpu_apply,
> >      },
> >  };
> >  
> > 
> 

-- 
David Gibson                    | I'll have my music baroque, and my code
david AT gibson.dropbear.id.au  | minimalist, thank you.  NOT _the_ _other_
                                | _way_ _around_!
http://www.ozlabs.org/~dgibson

signature.asc
Description: PGP signature