Bjoern Walk <bwalk@xxxxxxxxxxxxx> [2020-10-15, 09:41PM +0200]: > Jonathon Jongsma <jjongsma@xxxxxxxxxx> [2020-08-18, 09:47AM -0500]: > > diff --git a/tests/nodedevmdevctldata/mdevctl-list-single.out.xml b/tests/nodedevmdevctldata/mdevctl-list-single.out.xml > > new file mode 100644 > > index 0000000000..cb443346ef > > --- /dev/null > > +++ b/tests/nodedevmdevctldata/mdevctl-list-single.out.xml > > @@ -0,0 +1,14 @@ > > +<device> > > + <name>mdev_783e6dbb_ea0e_411f_94e2_717eaad438bf</name> > > + <parent>matrix</parent> > > + <capability type='mdev'> > > + <type id='vfio_ap-passthrough'/> > > + <iommuGroup number='0'/> > > + <attr name='assign_adapter' value='5'/> > > + <attr name='assign_adapter' value='6'/> > > + <attr name='assign_domain' value='0xab'/> > > + <attr name='assign_control_domain' value='0xab'/> > > + <attr name='assign_domain' value='4'/> > > + <attr name='assign_control_domain' value='4'/> > > + </capability> > > +</device> > > I find this XML design for the an AP mdev rather unfortunate. First of > all, it is really noisy and requires a lot of boiler-plate for the end > user. > > Secondly, which I think it more important, it introduces a completely > new and different nomenclature compared to what we currently already > have for AP-related devices, i.e. cards > > virsh # nodedev-dumpxml ap_card05 > <device> > <name>ap_card05</name> > <path>/sys/devices/ap/card05</path> > <parent>computer</parent> > <driver> > <name>cex4card</name> > </driver> > <capability type='ap_card'> > <ap-adapter>0x05</ap-adapter> > </capability> > </device> > > and queues > > virsh # nodedev-dumpxml ap_05_002a > <device> > <name>ap_05_002a</name> > <path>/sys/devices/ap/card05/05.002a</path> > <parent>ap_card05</parent> > <driver> > <name>cex4queue</name> > </driver> > <capability type='ap_queue'> > <ap-adapter>0x05</ap-adapter> > <ap-domain>0x002a</ap-domain> > </capability> > </device> I just figured that patches that implement that were not even sent upstream... We will do so shortly. I would however welcome a discussion on a sane and consistent XML design for both series. > > This is really inconsistent and can be confusing for the end user. > > Can we instead propose something along the lines > > <device> > <name>mdev_783e6dbb_ea0e_411f_94e2_717eaad438bf</name> > <parent>matrix</parent> > <capability type='mdev'> > <type id='vfio_ap-passthrough'/> > <iommuGroup number='0'/> > <ap-adapter>0x05</ap-adapter> > <ap-adapter>0x06</ap-adapter> > <ap-domain>0x002a</ap-domain> > <ap-domain>0x0004</ap-domain> > <ap-domain>0x00ab</ap-domain> > <ap-control>0x04</ap-control> > </capability> > </device> > > plus any number of enclosing elements that we like for better readability? > > Best, > Bjoern > > PS: I haven't read the whole patch series, we just have an internal > discussion about AP support and the XML design came up. I know that we'd > like to generically treat all mdevs as the same but we also should allow > for vendor-specific attributes to be represented individually to what > makes sense. > > -- > IBM Systems > Linux on Z & Virtualization Development > -------------------------------------------------- > IBM Deutschland Research & Development GmbH > Schönaicher Str. 220, 71032 Böblingen > Phone: +49 7031 16 1819 > -------------------------------------------------- > Vorsitzende des Aufsichtsrats: Gregor Pillen > Geschäftsführung: Dirk Wittkopp > Sitz der Gesellschaft: Böblingen > Registergericht: Amtsgericht Stuttgart, HRB 243294 -- IBM Systems Linux on Z & Virtualization Development -------------------------------------------------- IBM Deutschland Research & Development GmbH Schönaicher Str. 220, 71032 Böblingen Phone: +49 7031 16 1819 -------------------------------------------------- Vorsitzende des Aufsichtsrats: Gregor Pillen Geschäftsführung: Dirk Wittkopp Sitz der Gesellschaft: Böblingen Registergericht: Amtsgericht Stuttgart, HRB 243294
Attachment:
signature.asc
Description: PGP signature
