Re: [PATCH 2/9] iommu/arm-smmu: Calculate SMMU_CB_BASE from smmu register values

[Andreas Herrmann <andreas.herrmann-bsGFqQB8/DxBDgjK7y7TUQ@public.gmane.org>]

On Fri, Sep 27, 2013 at 06:51:53AM -0400, Will Deacon wrote:
> On Fri, Sep 27, 2013 at 11:23:59AM +0100, Andreas Herrmann wrote:
> > On Fri, Sep 27, 2013 at 05:51:57AM -0400, Will Deacon wrote:
> > > Hi Andreas,
> > > 
> > > On Thu, Sep 26, 2013 at 11:36:14PM +0100, Andreas Herrmann wrote:
> > > > Currently it is derived from smmu resource size.  In case of a
> > > > mismatchin between the two calculations trust DT more than register
> > > > values and overwrite cb_base.
> > > 
> > > I thought the driver already favoured the DT?
> > > 
> > > > @@ -1702,12 +1704,23 @@ static int arm_smmu_device_cfg_probe(struct arm_smmu_device *smmu)
> > > >  
> > > >  	/* Check that we ioremapped enough */
> > > >  	size = 1 << (((id >> ID1_NUMPAGENDXB_SHIFT) & ID1_NUMPAGENDXB_MASK) + 1);
> > > > -	size *= (smmu->pagesize << 1);
> > > > +	size *= smmu->pagesize;
> > > > +	smmu->cb_base = smmu->base + size;
> > > > +	size *= 2;
> > > > +
> > > >  	if (smmu->size < size)
> > > >  		dev_warn(smmu->dev,
> > > >  			 "device is 0x%lx bytes but only mapped 0x%lx!\n",
> > > >  			 size, smmu->size);
> > > >  
> > > > +	t = (unsigned long) smmu->base + (smmu->size >> 1);
> > > > +	if ((unsigned long)smmu->cb_base != t) {
> > > > +		dev_warn(smmu->dev, "address space mismatch, "
> > > > +			"overwriting cb_base (old: 0x%lx, new: 0x%lx)\n",
> > > > +			(unsigned long) smmu->cb_base, t);
> > > > +		smmu->cb_base = (void *) t;
> > > > +	}
> > > > +
> > > 
> > > I expect I'm just being slow here (only one coffee in), but I can't see what
> > > this gets us over the current use of resource_size (which goes and uses the
> > > DT).
> > 
> > On balance it adds a warning if there is an inconsistency between the
> > resource size and the relevant registers describing the SMMU address
> > space.
> 
> Well, we should already print the "device is 0x%lx bytes but only mapped
> 0x%lx!" message, which I think is enough to go and figure out what happened.

No, you can map a larger region and still wrongly calculate cb_base
w/o this warning.
So the required check is something like

         /* Check for mismatch between SMMU address space size and size of mapped region */
         size = 1 << (((id >> ID1_NUMPAGENDXB_SHIFT) & ID1_NUMPAGENDXB_MASK) + 1);
         size *= (smmu->pagesize << 1);
         if (smmu->size != size)
                 dev_warn(smmu->dev,
                          "SMMU_GLOBAL_SIZE (0x%lx) differs from mapped "
			  region size (0x%lx)!\n", size, smmu->size);



Andreas