RE: [PATCH v1 13/22] vfio: add bind stage-1 page table support

["Liu, Yi L" <yi.l.liu@intel.com>]

> From: Peter Xu <peterx@redhat.com>
> Sent: Wednesday, March 25, 2020 1:41 AM
> To: Liu, Yi L <yi.l.liu@intel.com>
> Subject: Re: [PATCH v1 13/22] vfio: add bind stage-1 page table support
> 
> On Sun, Mar 22, 2020 at 05:36:10AM -0700, Liu Yi L wrote:
> > This patch adds bind_stage1_pgtbl() definition in
> > HostIOMMUContextClass, also adds corresponding implementation in VFIO.
> > This is to expose a way for vIOMMU to setup dual stage DMA translation
> > for passthru devices on hardware.
> >
> > Cc: Kevin Tian <kevin.tian@intel.com>
> > Cc: Jacob Pan <jacob.jun.pan@linux.intel.com>
> > Cc: Peter Xu <peterx@redhat.com>
> > Cc: Eric Auger <eric.auger@redhat.com>
> > Cc: Yi Sun <yi.y.sun@linux.intel.com>
> > Cc: David Gibson <david@gibson.dropbear.id.au>
> > Cc: Alex Williamson <alex.williamson@redhat.com>
> > Signed-off-by: Liu Yi L <yi.l.liu@intel.com>
> > ---
> >  hw/iommu/host_iommu_context.c         | 49
> ++++++++++++++++++++++++++++++-
> >  hw/vfio/common.c                      | 55 ++++++++++++++++++++++++++++++++++-
> >  include/hw/iommu/host_iommu_context.h | 26 ++++++++++++++++-
> >  3 files changed, 127 insertions(+), 3 deletions(-)
> >
> > diff --git a/hw/iommu/host_iommu_context.c
> > b/hw/iommu/host_iommu_context.c index af61899..8a53376 100644
> > --- a/hw/iommu/host_iommu_context.c
> > +++ b/hw/iommu/host_iommu_context.c
> > @@ -69,21 +69,67 @@ int host_iommu_ctx_pasid_free(HostIOMMUContext
> *host_icx, uint32_t pasid)
> >      return hicxc->pasid_free(host_icx, pasid);  }
> >
> > +int host_iommu_ctx_bind_stage1_pgtbl(HostIOMMUContext *host_icx,
> > +                                     DualIOMMUStage1BindData *data) {
> > +    HostIOMMUContextClass *hicxc;
> > +
> > +    if (!host_icx) {
> > +        return -EINVAL;
> > +    }
> > +
> > +    hicxc = HOST_IOMMU_CONTEXT_GET_CLASS(host_icx);
> > +    if (!hicxc) {
> > +        return -EINVAL;
> > +    }
> > +
> > +    if (!(host_icx->flags & HOST_IOMMU_NESTING) ||
> > +        !hicxc->bind_stage1_pgtbl) {
> > +        return -EINVAL;
> > +    }
> > +
> > +    return hicxc->bind_stage1_pgtbl(host_icx, data); }
> > +
> > +int host_iommu_ctx_unbind_stage1_pgtbl(HostIOMMUContext *host_icx,
> > +                                       DualIOMMUStage1BindData *data)
> > +{
> > +    HostIOMMUContextClass *hicxc;
> > +
> > +    if (!host_icx) {
> > +        return -EINVAL;
> > +    }
> > +
> > +    hicxc = HOST_IOMMU_CONTEXT_GET_CLASS(host_icx);
> > +    if (!hicxc) {
> > +        return -EINVAL;
> > +    }
> > +
> > +    if (!(host_icx->flags & HOST_IOMMU_NESTING) ||
> > +        !hicxc->unbind_stage1_pgtbl) {
> > +        return -EINVAL;
> > +    }
> > +
> > +    return hicxc->unbind_stage1_pgtbl(host_icx, data); }
> > +
> >  void host_iommu_ctx_init(void *_host_icx, size_t instance_size,
> >                           const char *mrtypename,
> > -                         uint64_t flags)
> > +                         uint64_t flags, uint32_t formats)
> >  {
> >      HostIOMMUContext *host_icx;
> >
> >      object_initialize(_host_icx, instance_size, mrtypename);
> >      host_icx = HOST_IOMMU_CONTEXT(_host_icx);
> >      host_icx->flags = flags;
> > +    host_icx->stage1_formats = formats;
> >      host_icx->initialized = true;
> >  }
> >
> >  void host_iommu_ctx_destroy(HostIOMMUContext *host_icx)  {
> >      host_icx->flags = 0x0;
> > +    host_icx->stage1_formats = 0x0;
> 
> This could be dropped too with the function..

yes, it is.

> >      host_icx->initialized = false;
> >  }
> >
> > @@ -92,6 +138,7 @@ static void host_icx_init_fn(Object *obj)
> >      HostIOMMUContext *host_icx = HOST_IOMMU_CONTEXT(obj);
> >
> >      host_icx->flags = 0x0;
> > +    host_icx->stage1_formats = 0x0;
> 
> Same here...

echo.

> >      host_icx->initialized = false;
> >  }
> >
> > diff --git a/hw/vfio/common.c b/hw/vfio/common.c index
> > e0f2828..770a785 100644
> > --- a/hw/vfio/common.c
> > +++ b/hw/vfio/common.c
> > @@ -1223,6 +1223,52 @@ static int
> vfio_host_icx_pasid_free(HostIOMMUContext *host_icx,
> >      return 0;
> >  }
> >
> > +static int vfio_host_icx_bind_stage1_pgtbl(HostIOMMUContext
> > +*host_icx,
> 
> Same name issue on icx?  Feel free to choose anything that aligns with your
> previous decision...

yes, I'll use _host_iommu_ctx_ instead of _host_icx_.

> > +                                           DualIOMMUStage1BindData
> > +*bind_data) {
> > +    VFIOContainer *container = container_of(host_icx, VFIOContainer, host_icx);
> > +    struct vfio_iommu_type1_bind *bind;
> > +    unsigned long argsz;
> > +    int ret = 0;
> > +
> > +    argsz = sizeof(*bind) + sizeof(bind_data->bind_data);
> > +    bind = g_malloc0(argsz);
> > +    bind->argsz = argsz;
> > +    bind->flags = VFIO_IOMMU_BIND_GUEST_PGTBL;
> > +    memcpy(&bind->data, &bind_data->bind_data,
> > + sizeof(bind_data->bind_data));
> > +
> > +    if (ioctl(container->fd, VFIO_IOMMU_BIND, bind)) {
> > +        ret = -errno;
> > +        error_report("%s: pasid (%u) bind failed: %d",
> > +                      __func__, bind_data->pasid, ret);
> > +    }
> > +    g_free(bind);
> > +    return ret;
> > +}
> > +
> > +static int vfio_host_icx_unbind_stage1_pgtbl(HostIOMMUContext *host_icx,
> > +                                        DualIOMMUStage1BindData
> > +*bind_data) {
> > +    VFIOContainer *container = container_of(host_icx, VFIOContainer, host_icx);
> > +    struct vfio_iommu_type1_bind *bind;
> > +    unsigned long argsz;
> > +    int ret = 0;
> > +
> > +    argsz = sizeof(*bind) + sizeof(bind_data->bind_data);
> > +    bind = g_malloc0(argsz);
> > +    bind->argsz = argsz;
> > +    bind->flags = VFIO_IOMMU_UNBIND_GUEST_PGTBL;
> > +    memcpy(&bind->data, &bind_data->bind_data,
> > + sizeof(bind_data->bind_data));
> > +
> > +    if (ioctl(container->fd, VFIO_IOMMU_BIND, bind)) {
> > +        ret = -errno;
> > +        error_report("%s: pasid (%u) unbind failed: %d",
> > +                      __func__, bind_data->pasid, ret);
> > +    }
> > +    g_free(bind);
> > +    return ret;
> > +}
> > +
> >  /**
> >   * Get iommu info from host. Caller of this funcion should free
> >   * the memory pointed by the returned pointer stored in @info @@
> > -1337,6 +1383,7 @@ static int vfio_init_container(VFIOContainer *container, int
> group_fd,
> >          struct vfio_iommu_type1_info_cap_nesting nesting = {
> >                                           .nesting_capabilities = 0x0,
> >                                           .stage1_formats = 0, };
> > +        uint32_t stage1_formats;
> >
> >          ret = vfio_get_nesting_iommu_cap(container, &nesting);
> >          if (ret) {
> > @@ -1347,10 +1394,14 @@ static int vfio_init_container(VFIOContainer
> > *container, int group_fd,
> >
> >          flags |= (nesting.nesting_capabilities & VFIO_IOMMU_PASID_REQS) ?
> >                   HOST_IOMMU_PASID_REQUEST : 0;
> > +        flags |= HOST_IOMMU_NESTING;
> > +        stage1_formats = nesting.stage1_formats;
> > +
> >          host_iommu_ctx_init(&container->host_icx,
> >                              sizeof(container->host_icx),
> >                              TYPE_VFIO_HOST_IOMMU_CONTEXT,
> > -                            flags);
> > +                            flags,
> > +                            stage1_formats);
> 
> We can consider passing in nesting.stage1_formats and drop stage1_formats.

aha, yes.

Regards,
Yi Liu