Re: [Xen-devel] [PATCH v4 31/33] libxl: Add support for non-PCI passthrough

[Julien Grall <julien.grall@xxxxxxxxxx>]

Hi Ian,

On 31/03/15 12:49, Ian Campbell wrote:
> On Thu, 2015-03-19 at 19:29 +0000, Julien Grall wrote:
>> On ARM, every non-PCI device are described in the device tree. Each of
>> them can be found via a path.
>>
>> This patch introduces a very basic support, only the IOMMU will be set
>> up correctly. The user will have to:
>>     - Describe the device in the partial device tree
>>     - Map manually MMIO/IRQ
>>
>> This is a first approach, that will allow to have a basic non-PCI
>> passthrough support in Xen. This could be improved later.
>>
>> Furthermore add LIBXL_HAVE_DEVICETREE_PASSTHROUGH to indicate we
>> support non-PCI passthrough and partial device tree (introduced by a
>> previous patch).
> 
> Ah, you can ignore my comment on the previous patch then.
> 
> The comment with the #define might be an OK place for the
> big-scary-warning I mentioned in the previous patch too?

I think this IDL would be a more suitable place.

>>
>> Signed-off-by: Julien Grall <julien.grall@xxxxxxxxxx>
>> Cc: Ian Jackson <ian.jackson@xxxxxxxxxxxxx>
>> Cc: Wei Liu <wei.liu2@xxxxxxxxxx>
>>
>> ---
>>     Changes in v4:
>>         - Add LIBXL_HAVE_DEVICTREE_PASSTHROUGH to indicate we support
>>         non-PCI passthrough. This is also used in order to indicate
>>         partial device tree is supported
>>         - Remove libxl_dtdev.c as it contains only a 2 lines functions
>>         and call directly xc_* from libxl_create.c
>>         - Introduce domcreate_attach_dtdev
>>
>>     Changes in v3:
>>         - Dynamic allocation has been dropped
>>         - Rework the commit message in accordance with the previous
>>         item
>>
>>     Changes in v2:
>>         - Get DT infos earlier
>>         - Allocate/map IRQ in libxl__arch_domain_create rather than in
>>         libxl__device_dt_add
>>         - Use VIRQ rather than the PIRQ to construct the interrupts
>>         properties of the device tree
>>         - Correct cpumask in make_dtdev_node. We allow the interrupt to
>>         be used on the 8 CPUs
>>         - Fix LOGE when we map the MMIO region in the guest in
>>         libxl__device_dt_add. The domain and the IRQ were inverted
>>         - Calculate the number of SPIs to configure the VGIC
>>         - xc_physdev_dtdev_* helpers has been renamed to xc_dtdev_*
>>         - Rename libxl_device_dt to libxl_device_dtdev
>> ---
>>  tools/libxl/libxl.h          |  6 ++++++
>>  tools/libxl/libxl_create.c   | 32 ++++++++++++++++++++++++++++++++
>>  tools/libxl/libxl_internal.h |  5 +++++
>>  tools/libxl/libxl_types.idl  |  5 +++++
>>  4 files changed, 48 insertions(+)
>>
>> diff --git a/tools/libxl/libxl.h b/tools/libxl/libxl.h
>> index 6bc75c5..baaf06b 100644
>> --- a/tools/libxl/libxl.h
>> +++ b/tools/libxl/libxl.h
>> @@ -179,6 +179,12 @@
>>  #define LIBXL_HAVE_BUILDINFO_HVM_MMIO_HOLE_MEMKB 1
>>  
>>  /*
>> + * libxl_domain_build_info has device_tree and libxl_device_dtdev
>> + * exists. This mean non-PCI passthrough is supported for ARM
> 
> Rather than non-PCI I'd say device tree here, since I'm sure you aren't
> supporting the new flargle-bus I've just invented...

Ok

> 
>> + *
>> +#define LIBXL_HAVE_DEVICETREE_PASSTHROUGH 1
>> +
>> +/*
>>   * libxl ABI compatibility
>>   *
>>   * The only guarantee which libxl makes regarding ABI compatibility
>> diff --git a/tools/libxl/libxl_create.c b/tools/libxl/libxl_create.c
>> index 15b464e..39c828b 100644
>> --- a/tools/libxl/libxl_create.c
>> +++ b/tools/libxl/libxl_create.c
>> @@ -751,6 +751,8 @@ static void domcreate_attach_vtpms(libxl__egc *egc, 
>> libxl__multidev *multidev,
>>                                     int ret);
>>  static void domcreate_attach_pci(libxl__egc *egc, libxl__multidev *aodevs,
>>                                   int ret);
>> +static void domcreate_attach_dtdev(libxl__egc *egc,
>> +                                   libxl__domain_create_state *dcs);
>>  
>>  static void domcreate_console_available(libxl__egc *egc,
>>                                          libxl__domain_create_state *dcs);
>> @@ -1444,6 +1446,36 @@ static void domcreate_attach_pci(libxl__egc *egc, 
>> libxl__multidev *multidev,
>>          }
>>      }
>>  
>> +    domcreate_attach_dtdev(egc, dcs);
>> +    return;
>> +
>> +error_out:
>> +    assert(ret);
>> +    domcreate_complete(egc, dcs, ret);
>> +}
>> +
>> +static void domcreate_attach_dtdev(libxl__egc *egc,
>> +                                   libxl__domain_create_state *dcs)
> 
> I know it isn't strictly needed, but I think for consistency this
> function should take a ret and check it + propagate it via
> domcreate_complete on error, like the other ones in the call chain do.

The other caller have the ret because they use multidev. When it's not
the case (such as domcreate_console_available), the ret parameter is not
present.

So I'm not keen to add a new unused parameter.

Regards,

-- 
Julien Grall

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