On 8/2/19 3:30 PM, Peter Krempa wrote:
On Fri, Aug 02, 2019 at 15:15:59 +0200, Boris Fiuczynski wrote:
> Add support to specify a boot order on vfio-ccw passthrough devices.
>
> Signed-off-by: Boris Fiuczynski <fiuczy(a)linux.ibm.com>
> Acked-by: Jason J. Herne <jjherne(a)linux.ibm.com>
> Reviewed-by: Marc Hartmayer <mhartmay(a)linux.ibm.com>
> ---
> src/qemu/qemu_command.c | 27 +++++++++++++----
> .../hostdev-subsys-mdev-vfio-ccw-boot.args | 29 +++++++++++++++++++
> .../hostdev-subsys-mdev-vfio-ccw-boot.xml | 23 +++++++++++++++
> tests/qemuxml2argvtest.c | 4 +++
> 4 files changed, 78 insertions(+), 5 deletions(-)
> create mode 100644 tests/qemuxml2argvdata/hostdev-subsys-mdev-vfio-ccw-boot.args
> create mode 100644 tests/qemuxml2argvdata/hostdev-subsys-mdev-vfio-ccw-boot.xml
Note that I don't currently have enough information to verify that the
mdev claim is correct (e.g. why it's only with CCW devices?), but I have
two formal comments:
Since I know that on s390 you can boot from a vfio-ccw
mediated device
but not from a vfio-ap mediated device I chose the default as "not
capable to boot from" and check especially for the "known to be capable
to boot from" mdevs. I have no problem changing the logic if desired.
If we want to skip the checking than an qemu internal error like this
occurs:
error: Failed to start domain mini-ap
error: internal error: qemu unexpectedly closed the monitor:
2019-08-05T13:44:27.871697Z qemu-system-s390x: -device
vfio-ap,id=hostdev0,sysfsdev=/sys/bus/mdev/devices/9063cba3-ecef-47b6-abcf-3fef4fdcad85,bootindex=1:
Property '.bootindex' not found
>
> diff --git a/src/qemu/qemu_command.c b/src/qemu/qemu_command.c
> index fee51158a9..36138e2ccf 100644
> --- a/src/qemu/qemu_command.c
> +++ b/src/qemu/qemu_command.c
> @@ -5613,6 +5613,9 @@ qemuBuildHostdevMediatedDevStr(const virDomainDef *def,
> if (qemuBuildDeviceAddressStr(&buf, def, dev->info, qemuCaps) < 0)
> goto cleanup;
>
> + if (dev->info->bootIndex)
> + virBufferAsprintf(&buf, ",bootindex=%u",
dev->info->bootIndex);
> +
> if (virBufferCheckError(&buf) < 0)
> goto cleanup;
>
> @@ -5624,6 +5627,22 @@ qemuBuildHostdevMediatedDevStr(const virDomainDef *def,
> return ret;
> }
>
> +static bool
> +qemuHostdevSupportsBoot(virDomainHostdevDefPtr hostdev)
Refactor to this function should be separate from the logic change
below.
Agreed.
> +{
> + if (hostdev->mode == VIR_DOMAIN_HOSTDEV_MODE_SUBSYS) {
> + virDomainHostdevSubsysPtr subsys = &hostdev->source.subsys;
> + if (subsys->type == VIR_DOMAIN_HOSTDEV_SUBSYS_TYPE_PCI ||
> + subsys->type == VIR_DOMAIN_HOSTDEV_SUBSYS_TYPE_USB ||
> + subsys->type == VIR_DOMAIN_HOSTDEV_SUBSYS_TYPE_SCSI ||
> + (subsys->type == VIR_DOMAIN_HOSTDEV_SUBSYS_TYPE_MDEV &&
> + subsys->u.mdev.model == VIR_MDEV_MODEL_TYPE_VFIO_CCW)) {
This line is misaligned. Please make sure that it's clear that the above
line belongs to the term in the brackets.
I will fix it.
> + return true;
> + }
> + }
> + return false;
> +}
> +
> static int
> qemuBuildHostdevCommandLine(virCommandPtr cmd,
> const virDomainDef *def,
> @@ -5638,13 +5657,11 @@ qemuBuildHostdevCommandLine(virCommandPtr cmd,
> char *devstr;
>
> if (hostdev->info->bootIndex) {
> - if (hostdev->mode != VIR_DOMAIN_HOSTDEV_MODE_SUBSYS ||
> - (subsys->type != VIR_DOMAIN_HOSTDEV_SUBSYS_TYPE_PCI &&
> - subsys->type != VIR_DOMAIN_HOSTDEV_SUBSYS_TYPE_USB &&
> - subsys->type != VIR_DOMAIN_HOSTDEV_SUBSYS_TYPE_SCSI)) {
> + if (!qemuHostdevSupportsBoot(hostdev)) {
> virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
> _("booting from assigned devices is only
"
> - "supported for PCI, USB and SCSI
devices"));
> + "supported for PCI, USB, SCSI and MDEV "
> + "of model vfio-ccw devices"));
Ideally this validation will be done in the validation callback rather
than in the command line generator. (pre-existing so I don't require you
to refactor this at this time). >
Also this error can be probably stashed into qemuHostdevSupportsBoot.
(and name it like qemuHostdevValidateBoot).
Ok, I will rename the method and move the error report.
--
libvir-list mailing list
libvir-list(a)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: Matthias Hartmann
Geschäftsführung: Dirk Wittkopp
Sitz der Gesellschaft: Böblingen
Registergericht: Amtsgericht Stuttgart, HRB 243294