Re: [Qemu-devel] [PATCH 4/4] pci: add pci test device

["Michael S. Tsirkin" <mst@redhat.com>]

On Wed, Apr 03, 2013 at 02:05:32PM +0200, Paolo Bonzini wrote:
> Il 03/04/2013 14:00, Michael S. Tsirkin ha scritto:
> > On Wed, Apr 03, 2013 at 01:48:41PM +0200, Paolo Bonzini wrote:
> >> Il 03/04/2013 12:38, Michael S. Tsirkin ha scritto:
> >>> On Wed, Apr 03, 2013 at 12:34:24PM +0200, Paolo Bonzini wrote:
> >>>> Il 03/04/2013 12:33, Michael S. Tsirkin ha scritto:
> >>>>>> Because someone wasn't looking. :)  The hw/ reorganization patches I've
> >>>>>> posted fix that.
> >>>>>
> >>>>> Still not sure how to merge this, if it goes in through my tree
> >>>>> and that's the only comment, I'll just fix it silently ...
> >>>>
> >>>> Indeed.  In fact, considering my hw/ reorganization patches will be
> >>>> committed soon, please put it in hw/pci (matching hw/isa/pc-testdev.c).
> >>>
> >>> Paolo, hw/pci is pci core, I haven't looked at your reorg patches,
> >>> but please do not move devices there.
> >>> Sorting devices by connection is also wrong I think, by function
> >>> would be better.
> >>
> >> Indeed that's how most devices are sorted.  For example, PCI host
> >> devices/bridges/etc. are in hw/pci (together with the PCI core, making
> >> hw/pci basically all that goes in through your tree),
> > 
> > Well host bridges often do lots of things besides pci on the same
> > chip.
> 
> Still the PCIness is quite important, seeing that almost all such
> devices have a name that looks like *_pci.c.

Some of them do. For some of them name is not a very successful name.

> >> ISA host devices
> >> are in hw/isa, etc.
> > 
> > what do you mean "host ISA device"?
> 
> PCI-ISA bridges and SuperIO chips: piix4.c, vt82c686.c, lpc_ich9.c,
> i82378.c.  PIIX3 would also be there, it needs to be split out of
> piix_pci.c.
> 
> >> However, there are a few exceptions.  You cannot really sort out 600
> >> files without exceptions.  All USB devices are in hw/usb, and the
> >> existing test devices (debugexit, testdev) are in hw/isa.
> >>
> >> There is only one exception you should care about, namely that VFIO and
> >> ivshmem are also in hw/pci.
> > 
> > This makes no sense really.
> > Pls add hw/misc or just keep misc stuff in hw/
> 
> hw/misc exists already, I'll move those two there.
> 
> >> Here is the list of files in hw/pci:
> >>
> >> host-apb.c
> >> host-bonito.c
> >> host-dec.c
> >> host-dec.h
> >> host-grackle.c
> >> host-gt64xxx.c
> >> host-piix.c
> >> host-ppc4xx.c
> >> host-ppce500.c
> >> host-prep.c
> >> host-q35.c
> >> host-sh.c
> >> host-spapr.c
> >> host-uninorth.c
> >> host-versatile.c
> >> i82801b11.c
> >> ioh3420.c
> >> ioh3420.h
> >> ivshmem.c
> >> msi.c
> >> msi.h
> >> msix.c
> >> msix.h
> >> pam.c
> >> pci-hotplug.c
> >> pci-stub.c
> >> pci.c
> >> pci.h
> >> pci_bridge.c
> >> pci_bridge.h
> >> pci_bridge_dev.c
> >> pci_bus.h
> >> pci_host.c
> >> pci_host.h
> >> pci_ids.h
> >> pci_regs.h
> >> pcie.c
> >> pcie.h
> >> pcie_aer.c
> >> pcie_aer.h
> >> pcie_host.c
> >> pcie_host.h
> >> pcie_port.c
> >> pcie_port.h
> >> pcie_regs.h
> >> shpc.c
> >> shpc.h
> >> slotid_cap.c
> >> slotid_cap.h
> >> vfio.c
> >> xio3130_downstream.c
> >> xio3130_downstream.h
> >> xio3130_upstream.c
> >> xio3130_upstream.h
> >>
> >> Paolo
> > 
> > This messes up things.  pci core is separate from devices using it,
> > and it's important to me. Really just add hw/bridge/ and put all
> > kind of bridge devices there.
> 
> Ok, I'll add hw/pci/bridge, and remove the "host-" prefix for host PCI
> devices.
> 
> Paolo

That's too much nesting I think. hw/bridge/ and we can put isa bridges there
as well then.

-- 
MST