[PATCH 5/6 v5] kvm: booke: clear host tlb reference flag on guest tlb invalidation

[Bhushan Bharat-R65777]

> -----Original Message-----
> From: Wood Scott-B07421
> Sent: Friday, September 20, 2013 2:38 AM
> To: Bhushan Bharat-R65777
> Cc: benh at kernel.crashing.org; agraf at suse.de; paulus at samba.org;
> kvm at vger.kernel.org; kvm-ppc at vger.kernel.org; linuxppc-dev at lists.ozlabs.org;
> Bhushan Bharat-R65777
> Subject: Re: [PATCH 5/6 v5] kvm: booke: clear host tlb reference flag on guest
> tlb invalidation
> 
> On Thu, 2013-09-19 at 11:32 +0530, Bharat Bhushan wrote:
> > On booke, "struct tlbe_ref" contains host tlb mapping information
> > (pfn: for guest-pfn to pfn, flags: attribute associated with this
> > mapping) for a guest tlb entry. So when a guest creates a TLB entry
> > then "struct tlbe_ref" is set to point to valid "pfn" and set
> > attributes in "flags" field of the above said structure. When a guest
> > TLB entry is invalidated then flags field of corresponding "struct
> > tlbe_ref" is updated to point that this is no more valid, also we
> > selectively clear some other attribute bits, example: if
> > E500_TLB_BITMAP was set then we clear E500_TLB_BITMAP, if E500_TLB_TLB0 is set
> then we clear this.
> >
> > Ideally we should clear complete "flags" as this entry is invalid and
> > does not have anything to re-used. The other part of the problem is
> > that when we use the same entry again then also we do not clear (started doing
> or-ing etc).
> >
> > So far it was working because the selectively clearing mentioned above
> > actually clears "flags" what was set during TLB mapping. But the
> > problem starts coming when we add more attributes to this then we need
> > to selectively clear them and which is not needed.
> >
> > This patch we do both
> >         - Clear "flags" when invalidating;
> >         - Clear "flags" when reusing same entry later
> >
> > Signed-off-by: Bharat Bhushan <bharat.bhushan at freescale.com>
> > ---
> > v3-> v5
> >  - New patch (found this issue when doing vfio-pci development)
> >
> >  arch/powerpc/kvm/e500_mmu_host.c |   12 +++++++-----
> >  1 files changed, 7 insertions(+), 5 deletions(-)
> >
> > diff --git a/arch/powerpc/kvm/e500_mmu_host.c
> > b/arch/powerpc/kvm/e500_mmu_host.c
> > index 1c6a9d7..60f5a3c 100644
> > --- a/arch/powerpc/kvm/e500_mmu_host.c
> > +++ b/arch/powerpc/kvm/e500_mmu_host.c
> > @@ -217,7 +217,8 @@ void inval_gtlbe_on_host(struct kvmppc_vcpu_e500
> *vcpu_e500, int tlbsel,
> >  		}
> >  		mb();
> >  		vcpu_e500->g2h_tlb1_map[esel] = 0;
> > -		ref->flags &= ~(E500_TLB_BITMAP | E500_TLB_VALID);
> > +		/* Clear flags as TLB is not backed by the host anymore */
> > +		ref->flags = 0;
> >  		local_irq_restore(flags);
> >  	}
> 
> This breaks when you have both E500_TLB_BITMAP and E500_TLB_TLB0 set.

I do not see any case where we set both E500_TLB_BITMAP and E500_TLB_TLB0. Also we have not optimized that yet (keeping track of multiple shadow TLB0 entries for one guest TLB1 entry)

We uses these bit flags only for TLB1 and if size of stlbe is 4K then we set E500_TLB_TLB0  otherwise we set E500_TLB_BITMAP. Although I think that E500_TLB_BITMAP should be set only if stlbe size is less than gtlbe size.

> 
> Instead, just convert the final E500_TLB_VALID clearing at the end into
> ref->flags = 0, and convert the early return a few lines earlier into
> conditional execution of the tlbil_one().

This looks better, will send the patch shortly.

Thanks
-Bharat

> 
> -Scott
>