Re: [libvirt] [PATCH 0/8] vfio-ccw passthrough support

[Boris Fiuczynski]

On 05/07/2018 01:51 PM, John Ferlan wrote:

[...]


Someone may also want to consider creating a s390 specific version of
what Peter did for x86_64 for VIR_TEST_CAPS_LATEST in order to then
have/use the "latest" capabilities instead of adding bits to xml2argv
tests. I'm curious why the xml2xml test needed the bit adjustment - did
something fail?  Since there were no xml output data changes, that would
seem to indicate there isn't a need to modify the xml2xml test source.

I am not sure if I understood you correctly. Are you referring to patch
1? If so the changes are caused by a new QEMU_CAPS_CCW capability
replacing the QEMU_CAPS_VIRTIO_CCW capability. More is explained in the
commit message of the patch. In short: With support of vfio-ccw it
became apparent that the existence of the ccw bus is not well sourced by
observing virtio-ccw and therefore we replaced it with the detection of
the virtual-css-bridge. Let me know if I understood you wrong.



Sorry it wasn't clear enough - changes were recently made to
tests/qemuxml2argvtest.c in order to run the tests using the latest
capabilities rather than needing to pass each capability through the
test. The macros also have a version specific macro which allows for
checking/output from "previous" QEMU releases.

However, the changes only modified macros for x86_64 - so my comment was
if someone felt so inclined to avoid needing/checking specific/certain
caps and only cared that the latest caps did something a certain way,
then adjusting those macros/tests to handle s390* specific things would
perhaps helps achieve that. Furthermore, if previous QEMU versions would
produce different results, then using the version specific checks/output
would provide that support.


Oh, I tried to connect your comment to the vfio-ccw passthrough patch 
series but it seems rather unrelated to me.
Anyway you are right that it could be useful to extend the macros 
DO_TEST_CAPS_LATEST and DO_TEST_CAPS_VER beyond x86_64 arch. I do see 
these macros more like additional test scenarios that are a bit more 
reality oriented. The currently used approach of explicitly setting the 
minimum required capabilities for a single test case is still making 
sense to me as well since it allows to (slightly) narrow down the tested 
code path to the actual to be tested code/feature.



--
Mit freundlichen Grüßen/Kind regards
   Boris Fiuczynski

IBM Deutschland Research & Development GmbH
Vorsitzender des Aufsichtsrats: Martina Köderitz
Geschäftsführung: Dirk Wittkopp
Sitz der Gesellschaft: Böblingen
Registergericht: Amtsgericht Stuttgart, HRB 243294

--
libvir-list mailing list
libvir-list@redhat.com
https://www.redhat.com/mailman/listinfo/libvir-list

Re: [libvirt] [PATCH 0/8] vfio-ccw passthrough support

[John Ferlan]

[...]

>> Someone may also want to consider creating a s390 specific version of
>> what Peter did for x86_64 for VIR_TEST_CAPS_LATEST in order to then
>> have/use the "latest" capabilities instead of adding bits to xml2argv
>> tests. I'm curious why the xml2xml test needed the bit adjustment - did
>> something fail?  Since there were no xml output data changes, that would
>> seem to indicate there isn't a need to modify the xml2xml test source.
> I am not sure if I understood you correctly. Are you referring to patch
> 1? If so the changes are caused by a new QEMU_CAPS_CCW capability
> replacing the QEMU_CAPS_VIRTIO_CCW capability. More is explained in the
> commit message of the patch. In short: With support of vfio-ccw it
> became apparent that the existence of the ccw bus is not well sourced by
> observing virtio-ccw and therefore we replaced it with the detection of
> the virtual-css-bridge. Let me know if I understood you wrong.
> 

Sorry it wasn't clear enough - changes were recently made to
tests/qemuxml2argvtest.c in order to run the tests using the latest
capabilities rather than needing to pass each capability through the
test. The macros also have a version specific macro which allows for
checking/output from "previous" QEMU releases.

However, the changes only modified macros for x86_64 - so my comment was
if someone felt so inclined to avoid needing/checking specific/certain
caps and only cared that the latest caps did something a certain way,
then adjusting those macros/tests to handle s390* specific things would
perhaps helps achieve that. Furthermore, if previous QEMU versions would
produce different results, then using the version specific checks/output
would provide that support.

John

--
libvir-list mailing list
libvir-list@redhat.com
https://www.redhat.com/mailman/listinfo/libvir-list

Re: [libvirt] [PATCH 0/8] vfio-ccw passthrough support

[Boris Fiuczynski]

On 05/04/2018 05:07 PM, John Ferlan wrote:



On 04/26/2018 03:59 AM, Bjoern Walk wrote:

Shalini Chellathurai Saroja  [2018-04-11, 05:49PM 
+0200]:

Let us support the basic channel I/O passthrough infrastructure based on
vfio, which have been introduced in QEMU 2.10. The current focus is to
support dasd-eckd (cu_type/dev_type = 0x3990/0x3390) as the target
device.

Shalini Chellathurai Saroja (8):
   qemu: introduce capability for virtual-css-bridge
   qemu: introduce vfio-ccw capability
   util: virhostdev: add virHostdevIsMdevDevice()
   qemu: vfio-ccw device address generation
   qemu: command line generation for vfio-ccw device
   tests: tests for vfio-ccw passthrough
   docs: documentation for vfio-ccw passthrough
   news: documentation of new feature


Any chance, we get at least a review before 4.3 hits? Would be
appreciated.




So obviously this did not make the 4.3.0 and the series will need a
refresh due to the volume of change in qemu_capabilities.{c,h}.

I've reviewed a number of patches recently and made a similar comment in
all of them - when changing qemu_capabilities.{c,h} and updating the
various qemucapabilitiesdata/caps_*.xml files - do so in a separate
patch. That way if someone doesn't review the code right away, it's
actually fairly simple to recreate at least the capability for a
reviewer. Having it mixed in one patch with other qemu, conf, test, etc.
changes causes git am -3 to fail and thus makes review harder especially
when you don't get to reviews as soon as patches hit the list.

I am going to send out a new rebased version shortly.



Someone may also want to consider creating a s390 specific version of
what Peter did for x86_64 for VIR_TEST_CAPS_LATEST in order to then
have/use the "latest" capabilities instead of adding bits to xml2argv
tests. I'm curious why the xml2xml test needed the bit adjustment - did
something fail?  Since there were no xml output data changes, that would
seem to indicate there isn't a need to modify the xml2xml test source.
I am not sure if I understood you correctly. Are you referring to patch 
1? If so the changes are caused by a new QEMU_CAPS_CCW capability 
replacing the QEMU_CAPS_VIRTIO_CCW capability. More is explained in the 
commit message of the patch. In short: With support of vfio-ccw it 
became apparent that the existence of the ccw bus is not well sourced by 
observing virtio-ccw and therefore we replaced it with the detection of 
the virtual-css-bridge. Let me know if I understood you wrong.




John

--
libvir-list mailing list
libvir-list@redhat.com
https://www.redhat.com/mailman/listinfo/libvir-list




--
Mit freundlichen Grüßen/Kind regards
   Boris Fiuczynski

IBM Deutschland Research & Development GmbH
Vorsitzender des Aufsichtsrats: Martina Köderitz
Geschäftsführung: Dirk Wittkopp
Sitz der Gesellschaft: Böblingen
Registergericht: Amtsgericht Stuttgart, HRB 243294

--
libvir-list mailing list
libvir-list@redhat.com
https://www.redhat.com/mailman/listinfo/libvir-list

Re: [libvirt] [PATCH 0/8] vfio-ccw passthrough support

[John Ferlan]

On 04/26/2018 03:59 AM, Bjoern Walk wrote:
> Shalini Chellathurai Saroja  [2018-04-11, 05:49PM 
> +0200]:
>> Let us support the basic channel I/O passthrough infrastructure based on
>> vfio, which have been introduced in QEMU 2.10. The current focus is to
>> support dasd-eckd (cu_type/dev_type = 0x3990/0x3390) as the target
>> device.
>>
>> Shalini Chellathurai Saroja (8):
>>   qemu: introduce capability for virtual-css-bridge
>>   qemu: introduce vfio-ccw capability
>>   util: virhostdev: add virHostdevIsMdevDevice()
>>   qemu: vfio-ccw device address generation
>>   qemu: command line generation for vfio-ccw device
>>   tests: tests for vfio-ccw passthrough
>>   docs: documentation for vfio-ccw passthrough
>>   news: documentation of new feature
> 
> Any chance, we get at least a review before 4.3 hits? Would be
> appreciated.
> 
> 

So obviously this did not make the 4.3.0 and the series will need a
refresh due to the volume of change in qemu_capabilities.{c,h}.

I've reviewed a number of patches recently and made a similar comment in
all of them - when changing qemu_capabilities.{c,h} and updating the
various qemucapabilitiesdata/caps_*.xml files - do so in a separate
patch. That way if someone doesn't review the code right away, it's
actually fairly simple to recreate at least the capability for a
reviewer. Having it mixed in one patch with other qemu, conf, test, etc.
changes causes git am -3 to fail and thus makes review harder especially
when you don't get to reviews as soon as patches hit the list.

Someone may also want to consider creating a s390 specific version of
what Peter did for x86_64 for VIR_TEST_CAPS_LATEST in order to then
have/use the "latest" capabilities instead of adding bits to xml2argv
tests. I'm curious why the xml2xml test needed the bit adjustment - did
something fail?  Since there were no xml output data changes, that would
seem to indicate there isn't a need to modify the xml2xml test source.

John

--
libvir-list mailing list
libvir-list@redhat.com
https://www.redhat.com/mailman/listinfo/libvir-list

Re: [libvirt] [PATCH 0/8] vfio-ccw passthrough support

[Bjoern Walk]

Shalini Chellathurai Saroja  [2018-04-11, 05:49PM 
+0200]:
> Let us support the basic channel I/O passthrough infrastructure based on
> vfio, which have been introduced in QEMU 2.10. The current focus is to
> support dasd-eckd (cu_type/dev_type = 0x3990/0x3390) as the target
> device.
> 
> Shalini Chellathurai Saroja (8):
>   qemu: introduce capability for virtual-css-bridge
>   qemu: introduce vfio-ccw capability
>   util: virhostdev: add virHostdevIsMdevDevice()
>   qemu: vfio-ccw device address generation
>   qemu: command line generation for vfio-ccw device
>   tests: tests for vfio-ccw passthrough
>   docs: documentation for vfio-ccw passthrough
>   news: documentation of new feature

Any chance, we get at least a review before 4.3 hits? Would be
appreciated.


signature.asc
Description: PGP signature
--
libvir-list mailing list
libvir-list@redhat.com
https://www.redhat.com/mailman/listinfo/libvir-list

Re: [libvirt] [PATCH 0/8] vfio-ccw passthrough support

[Shalini Chellathurai Saroja]

Polite ping.


On 04/11/2018 05:49 PM, Shalini Chellathurai Saroja wrote:

Let us support the basic channel I/O passthrough infrastructure based on
vfio, which have been introduced in QEMU 2.10. The current focus is to
support dasd-eckd (cu_type/dev_type = 0x3990/0x3390) as the target
device.

Shalini Chellathurai Saroja (8):
   qemu: introduce capability for virtual-css-bridge
   qemu: introduce vfio-ccw capability
   util: virhostdev: add virHostdevIsMdevDevice()
   qemu: vfio-ccw device address generation
   qemu: command line generation for vfio-ccw device
   tests: tests for vfio-ccw passthrough
   docs: documentation for vfio-ccw passthrough
   news: documentation of new feature

  docs/drvnodedev.html.in|  21 -
  docs/formatdomain.html.in  |  20 +++-
  docs/news.xml  |   9 ++
  docs/schemas/domaincommon.rng  |   5 +-
  src/libvirt_private.syms   |   1 +
  src/qemu/qemu_capabilities.c   |  23 +
  src/qemu/qemu_capabilities.h   |   5 +
  src/qemu/qemu_command.c|  37 ++--
  src/qemu/qemu_domain.c |   2 +-
  src/qemu/qemu_domain_address.c |  32 +--
  src/qemu/qemu_hostdev.c|   3 +-
  src/qemu/qemu_hotplug.c|   4 +-
  src/util/virhostdev.c  |  26 --
  src/util/virhostdev.h  |   3 +
  src/util/virmdev.c |   3 +-
  src/util/virmdev.h |   1 +
  .../qemucapabilitiesdata/caps_2.10.0.s390x.replies |  28 --
  tests/qemucapabilitiesdata/caps_2.10.0.s390x.xml   |   3 +-
  .../qemucapabilitiesdata/caps_2.11.0.s390x.replies |  28 --
  tests/qemucapabilitiesdata/caps_2.11.0.s390x.xml   |   3 +-
  .../qemucapabilitiesdata/caps_2.12.0.s390x.replies |  31 +--
  tests/qemucapabilitiesdata/caps_2.12.0.s390x.xml   |   3 +
  .../qemucapabilitiesdata/caps_2.7.0.s390x.replies  |  24 +++--
  tests/qemucapabilitiesdata/caps_2.7.0.s390x.xml|   3 +-
  .../qemucapabilitiesdata/caps_2.8.0.s390x.replies  |  28 --
  tests/qemucapabilitiesdata/caps_2.8.0.s390x.xml|   3 +-
  .../qemucapabilitiesdata/caps_2.9.0.s390x.replies  |  28 --
  tests/qemucapabilitiesdata/caps_2.9.0.s390x.xml|   3 +-
  tests/qemuhotplugtest.c|   2 +-
  ...tdev-subsys-mdev-vfio-ccw-duplicate-address.xml |  29 ++
  ...ostdev-subsys-mdev-vfio-ccw-invalid-address.xml |  23 +
  .../hostdev-subsys-mdev-vfio-ccw.args  |  23 +
  .../hostdev-subsys-mdev-vfio-ccw.xml   |  22 +
  tests/qemuxml2argvtest.c   | 102 -
  .../hostdev-subsys-mdev-vfio-ccw.xml   |  28 ++
  tests/qemuxml2xmltest.c|  31 ---
  36 files changed, 491 insertions(+), 149 deletions(-)
  create mode 100644 
tests/qemuxml2argvdata/hostdev-subsys-mdev-vfio-ccw-duplicate-address.xml
  create mode 100644 
tests/qemuxml2argvdata/hostdev-subsys-mdev-vfio-ccw-invalid-address.xml
  create mode 100644 tests/qemuxml2argvdata/hostdev-subsys-mdev-vfio-ccw.args
  create mode 100644 tests/qemuxml2argvdata/hostdev-subsys-mdev-vfio-ccw.xml
  create mode 100644 tests/qemuxml2xmloutdata/hostdev-subsys-mdev-vfio-ccw.xml



--
libvir-list mailing list
libvir-list@redhat.com
https://www.redhat.com/mailman/listinfo/libvir-list

Re: [libvirt] [PATCH 0/8] vfio-ccw passthrough support

[Cornelia Huck]

On Wed, 11 Apr 2018 17:49:51 +0200
Shalini Chellathurai Saroja  wrote:

> Let us support the basic channel I/O passthrough infrastructure based on
> vfio, which have been introduced in QEMU 2.10. The current focus is to
> support dasd-eckd (cu_type/dev_type = 0x3990/0x3390) as the target
> device.

Neat.

Quick question: Does "current focus" mean "we have only tested it with
ECKD DASD", or "we only allow passthrough of 3990/3390 devices"?

> 
> Shalini Chellathurai Saroja (8):
>   qemu: introduce capability for virtual-css-bridge
>   qemu: introduce vfio-ccw capability
>   util: virhostdev: add virHostdevIsMdevDevice()
>   qemu: vfio-ccw device address generation
>   qemu: command line generation for vfio-ccw device
>   tests: tests for vfio-ccw passthrough
>   docs: documentation for vfio-ccw passthrough
>   news: documentation of new feature
> 
>  docs/drvnodedev.html.in|  21 -
>  docs/formatdomain.html.in  |  20 +++-
>  docs/news.xml  |   9 ++
>  docs/schemas/domaincommon.rng  |   5 +-
>  src/libvirt_private.syms   |   1 +
>  src/qemu/qemu_capabilities.c   |  23 +
>  src/qemu/qemu_capabilities.h   |   5 +
>  src/qemu/qemu_command.c|  37 ++--
>  src/qemu/qemu_domain.c |   2 +-
>  src/qemu/qemu_domain_address.c |  32 +--
>  src/qemu/qemu_hostdev.c|   3 +-
>  src/qemu/qemu_hotplug.c|   4 +-
>  src/util/virhostdev.c  |  26 --
>  src/util/virhostdev.h  |   3 +
>  src/util/virmdev.c |   3 +-
>  src/util/virmdev.h |   1 +
>  .../qemucapabilitiesdata/caps_2.10.0.s390x.replies |  28 --
>  tests/qemucapabilitiesdata/caps_2.10.0.s390x.xml   |   3 +-
>  .../qemucapabilitiesdata/caps_2.11.0.s390x.replies |  28 --
>  tests/qemucapabilitiesdata/caps_2.11.0.s390x.xml   |   3 +-
>  .../qemucapabilitiesdata/caps_2.12.0.s390x.replies |  31 +--
>  tests/qemucapabilitiesdata/caps_2.12.0.s390x.xml   |   3 +
>  .../qemucapabilitiesdata/caps_2.7.0.s390x.replies  |  24 +++--
>  tests/qemucapabilitiesdata/caps_2.7.0.s390x.xml|   3 +-
>  .../qemucapabilitiesdata/caps_2.8.0.s390x.replies  |  28 --
>  tests/qemucapabilitiesdata/caps_2.8.0.s390x.xml|   3 +-
>  .../qemucapabilitiesdata/caps_2.9.0.s390x.replies  |  28 --
>  tests/qemucapabilitiesdata/caps_2.9.0.s390x.xml|   3 +-
>  tests/qemuhotplugtest.c|   2 +-
>  ...tdev-subsys-mdev-vfio-ccw-duplicate-address.xml |  29 ++
>  ...ostdev-subsys-mdev-vfio-ccw-invalid-address.xml |  23 +
>  .../hostdev-subsys-mdev-vfio-ccw.args  |  23 +
>  .../hostdev-subsys-mdev-vfio-ccw.xml   |  22 +
>  tests/qemuxml2argvtest.c   | 102 
> -
>  .../hostdev-subsys-mdev-vfio-ccw.xml   |  28 ++
>  tests/qemuxml2xmltest.c|  31 ---
>  36 files changed, 491 insertions(+), 149 deletions(-)
>  create mode 100644 
> tests/qemuxml2argvdata/hostdev-subsys-mdev-vfio-ccw-duplicate-address.xml
>  create mode 100644 
> tests/qemuxml2argvdata/hostdev-subsys-mdev-vfio-ccw-invalid-address.xml
>  create mode 100644 tests/qemuxml2argvdata/hostdev-subsys-mdev-vfio-ccw.args
>  create mode 100644 tests/qemuxml2argvdata/hostdev-subsys-mdev-vfio-ccw.xml
>  create mode 100644 tests/qemuxml2xmloutdata/hostdev-subsys-mdev-vfio-ccw.xml
> 

--
libvir-list mailing list
libvir-list@redhat.com
https://www.redhat.com/mailman/listinfo/libvir-list

Re: [libvirt] [PATCH 0/8] vfio-ccw passthrough support

[Shalini Chellathurai Saroja]

On 04/11/2018 05:49 PM, Shalini Chellathurai Saroja wrote:

Let us support the basic channel I/O passthrough infrastructure based on
vfio, which have been introduced in QEMU 2.10. The current focus is to
support dasd-eckd (cu_type/dev_type = 0x3990/0x3390) as the target
device.


There is a temporary delay with our mail reception. The answer to the 
question below is as follows.


Quick question: Does "current focus" mean "we have only tested it with
ECKD DASD", or "we only allow passthrough of 3990/3390 devices"?

yes, it means we have only tested it with ECKD DASD device. The libvirt 
implementation supports all the devices which are supported by QEMU.


Thanks.



Shalini Chellathurai Saroja (8):
   qemu: introduce capability for virtual-css-bridge
   qemu: introduce vfio-ccw capability
   util: virhostdev: add virHostdevIsMdevDevice()
   qemu: vfio-ccw device address generation
   qemu: command line generation for vfio-ccw device
   tests: tests for vfio-ccw passthrough
   docs: documentation for vfio-ccw passthrough
   news: documentation of new feature

  docs/drvnodedev.html.in|  21 -
  docs/formatdomain.html.in  |  20 +++-
  docs/news.xml  |   9 ++
  docs/schemas/domaincommon.rng  |   5 +-
  src/libvirt_private.syms   |   1 +
  src/qemu/qemu_capabilities.c   |  23 +
  src/qemu/qemu_capabilities.h   |   5 +
  src/qemu/qemu_command.c|  37 ++--
  src/qemu/qemu_domain.c |   2 +-
  src/qemu/qemu_domain_address.c |  32 +--
  src/qemu/qemu_hostdev.c|   3 +-
  src/qemu/qemu_hotplug.c|   4 +-
  src/util/virhostdev.c  |  26 --
  src/util/virhostdev.h  |   3 +
  src/util/virmdev.c |   3 +-
  src/util/virmdev.h |   1 +
  .../qemucapabilitiesdata/caps_2.10.0.s390x.replies |  28 --
  tests/qemucapabilitiesdata/caps_2.10.0.s390x.xml   |   3 +-
  .../qemucapabilitiesdata/caps_2.11.0.s390x.replies |  28 --
  tests/qemucapabilitiesdata/caps_2.11.0.s390x.xml   |   3 +-
  .../qemucapabilitiesdata/caps_2.12.0.s390x.replies |  31 +--
  tests/qemucapabilitiesdata/caps_2.12.0.s390x.xml   |   3 +
  .../qemucapabilitiesdata/caps_2.7.0.s390x.replies  |  24 +++--
  tests/qemucapabilitiesdata/caps_2.7.0.s390x.xml|   3 +-
  .../qemucapabilitiesdata/caps_2.8.0.s390x.replies  |  28 --
  tests/qemucapabilitiesdata/caps_2.8.0.s390x.xml|   3 +-
  .../qemucapabilitiesdata/caps_2.9.0.s390x.replies  |  28 --
  tests/qemucapabilitiesdata/caps_2.9.0.s390x.xml|   3 +-
  tests/qemuhotplugtest.c|   2 +-
  ...tdev-subsys-mdev-vfio-ccw-duplicate-address.xml |  29 ++
  ...ostdev-subsys-mdev-vfio-ccw-invalid-address.xml |  23 +
  .../hostdev-subsys-mdev-vfio-ccw.args  |  23 +
  .../hostdev-subsys-mdev-vfio-ccw.xml   |  22 +
  tests/qemuxml2argvtest.c   | 102 -
  .../hostdev-subsys-mdev-vfio-ccw.xml   |  28 ++
  tests/qemuxml2xmltest.c|  31 ---
  36 files changed, 491 insertions(+), 149 deletions(-)
  create mode 100644 
tests/qemuxml2argvdata/hostdev-subsys-mdev-vfio-ccw-duplicate-address.xml
  create mode 100644 
tests/qemuxml2argvdata/hostdev-subsys-mdev-vfio-ccw-invalid-address.xml
  create mode 100644 tests/qemuxml2argvdata/hostdev-subsys-mdev-vfio-ccw.args
  create mode 100644 tests/qemuxml2argvdata/hostdev-subsys-mdev-vfio-ccw.xml
  create mode 100644 tests/qemuxml2xmloutdata/hostdev-subsys-mdev-vfio-ccw.xml



--
libvir-list mailing list
libvir-list@redhat.com
https://www.redhat.com/mailman/listinfo/libvir-list