Re: [Qemu-devel] [PATCH 01/15] s390 vfio-ccw: Add bootindex property and IPLB data

["Jason J. Herne" <jjherne@linux.ibm.com>]

On 2/4/19 5:26 AM, Cornelia Huck wrote:
> On Tue, 29 Jan 2019 08:29:08 -0500
> "Jason J. Herne" <jjherne@linux.ibm.com> wrote:
> 
>> Add bootindex property and iplb data for vfio-ccw devices. This allows us to
>> forward boot information into the bios for vfio-ccw devices.
>>
>> Signed-off-by: Jason J. Herne <jjherne@linux.ibm.com>
>> Acked-by: Halil Pasic <pasic@linux.vnet.ibm.com>
>> ---
>>   hw/s390x/ipl.c              | 14 ++++++++++++++
>>   hw/s390x/s390-ccw.c         |  9 +++++++++
>>   hw/vfio/ccw.c               | 13 +------------
>>   include/hw/s390x/s390-ccw.h |  1 +
>>   include/hw/s390x/vfio-ccw.h | 38 ++++++++++++++++++++++++++++++++++++++
>>   5 files changed, 63 insertions(+), 12 deletions(-)
>>   create mode 100644 include/hw/s390x/vfio-ccw.h
>>
>> diff --git a/hw/s390x/ipl.c b/hw/s390x/ipl.c
>> index 21f64ad..a993f65 100644
>> --- a/hw/s390x/ipl.c
>> +++ b/hw/s390x/ipl.c
>> @@ -19,6 +19,7 @@
>>   #include "hw/loader.h"
>>   #include "hw/boards.h"
>>   #include "hw/s390x/virtio-ccw.h"
>> +#include "hw/s390x/vfio-ccw.h"
>>   #include "hw/s390x/css.h"
>>   #include "hw/s390x/ebcdic.h"
>>   #include "ipl.h"
>> @@ -311,8 +312,12 @@ static CcwDevice *s390_get_ccw_device(DeviceState *dev_st)
>>           VirtioCcwDevice *virtio_ccw_dev = (VirtioCcwDevice *)
>>               object_dynamic_cast(OBJECT(qdev_get_parent_bus(dev_st)->parent),
>>                                   TYPE_VIRTIO_CCW_DEVICE);
>> +        VFIOCCWDevice *vfio_ccw_dev = (VFIOCCWDevice *)
>> +            object_dynamic_cast(OBJECT(dev_st), TYPE_VFIO_CCW);
>>           if (virtio_ccw_dev) {
>>               ccw_dev = CCW_DEVICE(virtio_ccw_dev);
>> +        } else if (vfio_ccw_dev) {
>> +            ccw_dev = CCW_DEVICE(vfio_ccw_dev);
>>           } else {
>>               SCSIDevice *sd = (SCSIDevice *)
>>                   object_dynamic_cast(OBJECT(dev_st),
>> @@ -347,6 +352,8 @@ static bool s390_gen_initial_iplb(S390IPLState *ipl)
>>       if (ccw_dev) {
>>           SCSIDevice *sd = (SCSIDevice *) object_dynamic_cast(OBJECT(dev_st),
>>                                                               TYPE_SCSI_DEVICE);
>> +        VFIOCCWDevice *vc = (VFIOCCWDevice *)
>> +            object_dynamic_cast(OBJECT(dev_st), TYPE_VFIO_CCW);
>>   
>>           if (sd) {
>>               ipl->iplb.len = cpu_to_be32(S390_IPLB_MIN_QEMU_SCSI_LEN);
>> @@ -358,6 +365,13 @@ static bool s390_gen_initial_iplb(S390IPLState *ipl)
>>               ipl->iplb.scsi.channel = cpu_to_be16(sd->channel);
>>               ipl->iplb.scsi.devno = cpu_to_be16(ccw_dev->sch->devno);
>>               ipl->iplb.scsi.ssid = ccw_dev->sch->ssid & 3;
>> +        } else if (vc) {
>> +            CcwDevice *ccw_dev = CCW_DEVICE(vc);
>> +
>> +            ipl->iplb.len = cpu_to_be32(S390_IPLB_MIN_CCW_LEN);
>> +            ipl->iplb.pbt = S390_IPL_TYPE_CCW;
>> +            ipl->iplb.ccw.devno = cpu_to_be16(ccw_dev->sch->devno);
>> +            ipl->iplb.ccw.ssid = ccw_dev->sch->ssid & 3;
>>           } else {
>>               VirtIONet *vn = (VirtIONet *) object_dynamic_cast(OBJECT(dev_st),
>>                                                                 TYPE_VIRTIO_NET);
> 
> Hm, I think that this find-out-the-boot-type-and-set-up-the-right-thing
> mechanism is getting a bit unwieldy. Basically, we
> 
> - call s390_get_ccw_device() to find out the device type via a bunch of
>    casts and return a pointer to a CcwDevice if it's a supported type
> - do a bunch of casts here *again* to find out what we have and fill
>    out the iplb, while we really only need to do grab a non-CcwDevice
>    for the scsi case
> 
> Should maybe s390_get_ccw_device() give us an ipl type in addition to
> the pointer to the CcwDevice, so we can use a switch/case statement to
> fill out the iplb here?

I think this idea makes sense. s390_ipl_reset_request also calls s390_get_ccw_device but 
does not care bout the device type, so how about a separate function instead of 
integrating it with s390_get_ccw_device? Maybe s390_get_ccw_device_type?

Is there any easy way to grab the type or are we just hiding the ugly casting inside our 
helper function?


-- 
-- Jason J. Herne (jjherne@linux.ibm.com)