RE: [PATCH v2 4/5] iommu: Support mm PASID 1:n with sva domains

[Zhang, Tina]

Hi Vasant,

> -----Original Message-----
> From: Vasant Hegde <vasant.hegde@xxxxxxx>
> Sent: Monday, August 28, 2023 4:33 PM
> To: Zhang, Tina <tina.zhang@xxxxxxxxx>; Jason Gunthorpe <jgg@xxxxxxxx>;
> Tian, Kevin <kevin.tian@xxxxxxxxx>; Lu Baolu <baolu.lu@xxxxxxxxxxxxxxx>;
> Michael Shavit <mshavit@xxxxxxxxxx>
> Cc: iommu@xxxxxxxxxxxxxxx; linux-kernel@xxxxxxxxxxxxxxx
> Subject: Re: [PATCH v2 4/5] iommu: Support mm PASID 1:n with sva domains
> 
> Hi Tina,
> 
> On 8/27/2023 2:14 PM, Tina Zhang wrote:
> > Each mm bound to devices gets a PASID and corresponding sva domains
> > allocated in iommu_sva_bind_device(), which are referenced by
> iommu_mm
> > field of the mm. The PASID is released in __mmdrop(), while a sva
> > domain is released when no one is using it (the reference count is
> > decremented in iommu_sva_unbind_device()).
> >
> > Since the required info of PASID and sva domains is kept in struct
> > iommu_mm_data of a mm, use mm->iommu_mm field instead of the old
> pasid
> > field in mm struct. The sva domain list is protected by iommu_sva_lock.
> >
> > Besides, this patch removes mm_pasid_init(), as with the introduced
> > iommu_mm structure, initializing mm pasid in mm_init() is unnecessary.
> >
> > Signed-off-by: Tina Zhang <tina.zhang@xxxxxxxxx>
> > ---
> >  drivers/iommu/iommu-sva.c | 38 +++++++++++++++++++++++++-------------
> >  include/linux/iommu.h     | 10 +++-------
> >  kernel/fork.c             |  1 -
> >  3 files changed, 28 insertions(+), 21 deletions(-)
> >
> 
> 
> .../...
> 
> >
> >  	/* Allocate a new domain and set it on device pasid. */ @@ -105,6
> > +113,8 @@ struct iommu_sva *iommu_sva_bind_device(struct device *dev,
> struct mm_struct *mm
> >  	if (ret)
> >  		goto out_free_domain;
> >  	domain->users = 1;
> > +	list_add(&domain->next, &mm->iommu_mm->sva_domains);
> > +
> >  out:
> >  	mutex_unlock(&iommu_sva_lock);
> >  	handle->dev = dev;
> > @@ -137,8 +147,9 @@ void iommu_sva_unbind_device(struct iommu_sva
> *handle)
> >  	struct device *dev = handle->dev;
> >
> >  	mutex_lock(&iommu_sva_lock);
> > +	iommu_detach_device_pasid(domain, dev, pasid);
> >  	if (--domain->users == 0) {
> > -		iommu_detach_device_pasid(domain, dev, pasid);
> > +		list_del(&domain->next);
> >  		iommu_domain_free(domain);
> >  	}
> >  	mutex_unlock(&iommu_sva_lock);
> > @@ -218,4 +229,5 @@ void mm_pasid_drop(struct mm_struct *mm)
> >  		return;
> >
> >  	ida_free(&iommu_global_pasid_ida, mm_get_pasid(mm));
> > +	kfree(mm->iommu_mm);
> 
> 
> I am not sure whether I understood the flow completely. Just wondering why
> you are not freeing pasid in iommu_sva_unbind_device().
> I mean once iommu_mm->sva_domains becomes free shouldn't we free the
> iommu_mm->pasid?
No, the sva domain and the PASID are separate objects with their own lifecycles.
The iommu_mm->pasid is released when the mm is being released, meanwhile the sva_domain is released when no one is using it.

Regards,
-Tina
> 
> Also in this function (mm_pasid_drop()), should we check/free sva domains?
> 
> -Vasant