Re: [PATCH 2/7] xen/arm: SMMU: Introduce a helper to add DT device to SMMU

[Wei Chen <Wei.Chen@arm.com>]

Hi Stefano,

> -----Original Message-----
> From: Stefano Stabellini [mailto:sstabellini@kernel.org]
> Sent: 2017年7月4日 6:03
> To: Wei Chen <Wei.Chen@arm.com>
> Cc: xen-devel@lists.xen.org; sstabellini@kernel.org; Steve Capper
> <Steve.Capper@arm.com>; Kaly Xin <Kaly.Xin@arm.com>; Julien Grall
> <Julien.Grall@arm.com>; nd <nd@arm.com>
> Subject: Re: [Xen-devel] [PATCH 2/7] xen/arm: SMMU: Introduce a helper to add
> DT device to SMMU
> 
> On Fri, 30 Jun 2017, Wei Chen wrote:
> > In current code, we only have the iommu_add_device to add PCI device
> > to IOMMU. But for ARM SMMU, we don't have a separate helper to add
> > platform device with device tree to SMMU. This work was included in
> > the iommu_assign_dt_device. But sometimes, we just want to add device
> > to SMMU to do some preparation for further use. In this case, we can't
> > call iommu_assign_dt_device.
> >
> > In previous patch, we have implement the add_device callback for SMMU,
> > so we can separate this work from assign_device now.
> >
> > Signed-off-by: Wei Chen <Wei.Chen@arm.com>
> > ---
> >  xen/drivers/passthrough/device_tree.c | 20 ++++++++++++++++++++
> >  xen/include/xen/iommu.h               |  1 +
> >  2 files changed, 21 insertions(+)
> >
> > diff --git a/xen/drivers/passthrough/device_tree.c
> b/xen/drivers/passthrough/device_tree.c
> > index 99ed49e..a8f403a 100644
> > --- a/xen/drivers/passthrough/device_tree.c
> > +++ b/xen/drivers/passthrough/device_tree.c
> > @@ -24,6 +24,26 @@
> >
> >  static spinlock_t dtdevs_lock = SPIN_LOCK_UNLOCKED;
> >
> > +int iommu_add_dt_device(struct domain *d, struct dt_device_node *dev)
> > +{
> > +    int rc;
> > +
> > +    struct domain_iommu *hd = dom_iommu(d);
> > +
> > +    if ( !iommu_enabled || !hd->platform_ops ||
> > +         !hd->platform_ops->add_device )
> > +        return 0;
> 
> Shouldn't we also have:
> 
>   if ( !dt_device_is_protected(dev) )
>         return 0;
> 
> ?
> 

When we're using the legacy binding, the master IDs will be registered to SMMU and
the protected flag of relevant master devices will be set to true (dt_device_is_protected
will return true). But for generic IOMMU bindings, before we call ops->add_device,
we didn't register the master device's master id to SMMU and hadn't set the protected
flag, The dt_device_is_protected will always return false.

In this case, we can't call dt_device_is_protected(dev) here.

> 
> > +    spin_lock(&dtdevs_lock);
> > +
> > +    /* The devfn field doesn't matter to DT device. */
> > +    rc = hd->platform_ops->add_device(0, dt_to_dev(dev));
> > +
> > +    spin_unlock(&dtdevs_lock);
> > +
> > +    return rc;
> > +}
> > +
> >  int iommu_assign_dt_device(struct domain *d, struct dt_device_node *dev)
> >  {
> >      int rc = -EBUSY;
> > diff --git a/xen/include/xen/iommu.h b/xen/include/xen/iommu.h
> > index 5803e3f..ec03faa 100644
> > --- a/xen/include/xen/iommu.h
> > +++ b/xen/include/xen/iommu.h
> > @@ -132,6 +132,7 @@ void iommu_read_msi_from_ire(struct msi_desc *msi_desc,
> struct msi_msg *msg);
> >  #ifdef CONFIG_HAS_DEVICE_TREE
> >  #include <xen/device_tree.h>
> >
> > +int iommu_add_dt_device(struct domain *d, struct dt_device_node *dev);
> >  int iommu_assign_dt_device(struct domain *d, struct dt_device_node *dev);
> >  int iommu_deassign_dt_device(struct domain *d, struct dt_device_node *dev);
> >  int iommu_dt_domain_init(struct domain *d);
> > --
> > 2.7.4
> >
> >
> > _______________________________________________
> > Xen-devel mailing list
> > Xen-devel@lists.xen.org
> > https://lists.xen.org/xen-devel
> >

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
https://lists.xen.org/xen-devel