Re: [PATCH v2 05/11] iommu/sva: Assign a PASID to mm on PASID allocation and free it on mm exit

[Fenghua Yu <fenghua.yu@intel.com>]

Hi, Thomas,

On Wed, Jan 26, 2022 at 10:38:04PM +0100, Thomas Gleixner wrote:
> On Wed, Jan 26 2022 at 09:36, Fenghua Yu wrote:
> > On Wed, Jan 26, 2022 at 03:23:42PM +0100, Thomas Gleixner wrote:
> >> On Tue, Jan 25 2022 at 07:18, Fenghua Yu wrote:
> >> While looking at ioasid_put() usage I tripped over the following UAF
> >> issue:
> >> 
> >> --- a/drivers/iommu/intel/iommu.c
> >> +++ b/drivers/iommu/intel/iommu.c
> >> @@ -4817,8 +4817,10 @@ static int aux_domain_add_dev(struct dma
> >>  	auxiliary_unlink_device(domain, dev);
> >>  link_failed:
> >>  	spin_unlock_irqrestore(&device_domain_lock, flags);
> >> -	if (list_empty(&domain->subdevices) && domain->default_pasid > 0)
> >> +	if (list_empty(&domain->subdevices) && domain->default_pasid > 0) {
> >>  		ioasid_put(domain->default_pasid);
> >> +		domain->default_pasid = INVALID_IOASID;
> >> +	}
> >>  
> >>  	return ret;
> >>  }
> >> @@ -4847,8 +4849,10 @@ static void aux_domain_remove_dev(struct
> >>  
> >>  	spin_unlock_irqrestore(&device_domain_lock, flags);
> >>  
> >> -	if (list_empty(&domain->subdevices) && domain->default_pasid > 0)
> >> +	if (list_empty(&domain->subdevices) && domain->default_pasid > 0) {
> >>  		ioasid_put(domain->default_pasid);
> >> +		domain->default_pasid = INVALID_IOASID;
> >> +	}
> >>  }
> >>  
> >>  static int prepare_domain_attach_device(struct iommu_domain *domain,
> >
> > The above patch fixes an existing issue. I will put it in a separate patch,
> > right?
> 
> Correct.
> 
> > It cannot be applied cleanly to the upstream tree. Do you want me to base
> > the above patch (and the whole patch set) to the upstream tree or a specific
> > tip branch?
> 
> Against Linus tree please so that the bugfix applies.
> 
> > I will fold the following patch into patch #5. The patch #11 (the doc patch)
> > also needs to remove one paragraph talking about refcount.
> >
> > So I will send the whole patch set with the following changes:
> > 1. One new bug fix patch (the above patch)

When I study your above aux_domain bug fix path, I find more aux_domain bugs.
But then I find aux_domain will be completely removed because all aux_domain
related callbacks are not called and are dead code (no wonder there are
so many bugs in aux_domain). Please see this series: https://lore.kernel.org/linux-iommu/20220124071103.2097118-4-baolu.lu@linux.intel.com/
For the series, Baolu confirms that he is "pretty sure that should be part
of v5.18". And I don't find the series calls any IOASID function after
removing the aux_domain code.

So that means we don't need to fix those issues in the dead aux_domain
code any more because it will be completely removed in 5.18, right?

If you agree, I will not include the aux_domain fix patch or any other
aux_domain fix patches in the up-coming v3. Is that OK?

> > 2. Updated patch #5 (with the following patch folded)

I will still change ioasid_put() in the aux_domain callbacks to ioasid_free()
in patch #5. So compilation will pass. Baolu's series will remove
the entire aux_domain code in 5.18.

> > 3. Updated patch #11 (removing refcount description)
> 
> Looks good.
> 

Thanks.

-Fenghua