[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Qemu-ppc] [Qemu-devel] [PATCH 04/12] spapr_pci: add set-indicator R

From: Michael Roth
Subject: Re: [Qemu-ppc] [Qemu-devel] [PATCH 04/12] spapr_pci: add set-indicator RTAS interface
Date: Tue, 23 Dec 2014 09:12:39 -0600
User-agent: alot/0.3.4

Quoting Bharata B Rao (2014-11-30 22:57:48)
> On Wed, Nov 26, 2014 at 11:57 AM, Michael Roth
> <address@hidden> wrote:
> > https://github.com/mdroth/qemu/commits/spapr-pci-hotplug-ppc-next-cleanup4.2
> >
> > The sPAPRDREntry stuff is now modeled by the sPAPRDRConnector QOM object in
> > hw/ppc/spapr_drc.c, which manages the device's life-cycle based on
> > rtas-set-sensor-state calls from the guest. As part of qemu-side 
> > hotplug/unplug
> > you use the attach/detach methods of the DRC to associate DT bits and 
> > callbacks
> > for things like device cleanup or rtas calls to fetch a DT node from the 
> > device
> > associated with a particular DRC.
> >
> > I still need to fix endian issues, and am realizing the dr connectors and DT
> > bits for PHBs are not actually a prereq for PCI hotplug, so I may be pulling
> > that out to a separate series specific to enabling PHB hotplug (namely for
> > VFIO hotplug). I realize your CPU/MEM sort of depend on the top-level PHB
> > device tree code so I'm not sure how best to deal with that. Worse case we'd
> > roll the initial code into your series and base a follow-up series on that 
> > of
> > that instead.
> Thanks Michael for pointing me to your git tree.
> I started rebasing my patchset on top of yours and realized that the
> generic DT setup code from the below commits of your branch are needed
> for CPU and memory hotplug too. They all apply in the order I  have
> listed below.
> 71b32999c4eb spapr_drc: initial implementation
> 255c50200848 spapr: populate DRC entries for root dt node (don't need
> code that adds PHB DT entries)
> 408206fc627e3 spapr_rtas: add set-indicator RTAS interface
> da7a232fa6a44 spapr_rtas: add get-sensor-state RTAS interface
> 1c575d5b29688 spapr_rtas: add ibm,configure-connector RTAS interface
> 0c5d72833666c spapr_events: re-use EPOW event infrastructure for hotplug 
> events
> 82ee5a9c88155 spapr_events: event-scan RTAS interface
> If you can make the above set an independent patchset, it will become
> easy to maintain and post CPU and memory hotplug patchsets.

Hi Bharata,

I've submitted v4 of PCI hotplug. The development branch is here:


and is based on top of a 'core' branch organized similar to what you proposed:


I'll be rolling changes for core/pci code into the branches as we go.
The endian fixes you provided are included, and PCI hotplug has been
tested on ppc64le.

There's a pseries-2.3 in the core patchset to enable/disable
dynamic-reconfiguration for individual resources on a machine basis to
maintain backward migration compatibility. There's a PHB hotplug patchset
based on core that might be a good reference for re-basing CPU/memory:


> I am facing some endian issues in your patchset and I will send fixes
> for those separately.
> Regards,
> Bharata.

reply via email to

[Prev in Thread] Current Thread [Next in Thread]