[Qemu-devel] Re: [PATCH 2/2] pci: don't overwrite pci header type.

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

On Wed, Jun 16, 2010 at 06:43:53PM +0900, Isaku Yamahata wrote:
> On Wed, Jun 16, 2010 at 11:54:25AM +0300, Michael S. Tsirkin wrote:
> > On Wed, Jun 16, 2010 at 11:20:02AM +0900, Isaku Yamahata wrote:
> > > On Tue, Jun 15, 2010 at 12:12:07PM +0300, Michael S. Tsirkin wrote:
> > > > On Tue, Jun 15, 2010 at 02:06:46PM +0900, Isaku Yamahata wrote:
> > > > > Don't overwrite pci header type.
> > > > > Otherwise, multi function bit which pci_init_header_type() sets
> > > > > appropriately is lost.
> > > > > Anyway PCI_HEADER_TYPE_NORMAL is zero, so it is unnecessary to zero
> > > > > which is already zero cleared.
> > > > > 
> > > > > Signed-off-by: Isaku Yamahata <yamahata@valinux.co.jp>
> > > > 
> > > > ...
> > > > 
> > > > > diff --git a/hw/apb_pci.c b/hw/apb_pci.c
> > > > > index 31c8d70..cdf3bc2 100644
> > > > > --- a/hw/apb_pci.c
> > > > > +++ b/hw/apb_pci.c
> > > > > @@ -428,7 +428,8 @@ static int pbm_pci_host_init(PCIDevice *d)
> > > > >                   PCI_STATUS_DEVSEL_MEDIUM);
> > > > >      pci_config_set_class(d->config, PCI_CLASS_BRIDGE_HOST);
> > > > >      pci_set_byte(d->config + PCI_HEADER_TYPE,
> > > > > -                 PCI_HEADER_TYPE_NORMAL);
> > > > > +                 (pci_get_byte(d->config + PCI_HEADER_TYPE) &
> > > > > +                  PCI_HEADER_TYPE_MULTI_FUNCTION) | PCI_HEADER_TYPE_NORMAL);
> > > > 
> > > > what is this doing?
> > > 
> > > It changes the header type to normal device(bit 1-7) without overwriting
> > > multi function bit(bit 8).
> > 
> > Don't we know what the multi function bit value is?
> 
> pci generic initialization, pci_qdev_init(), in pci.c sets (or clears) the bit
> and then calls the device specific initialization function, pbm_pci_host_init()
> in this case.
> So we shouldn't clear the bit unconditionally.
> 
> 
> > > Apb host bridge specifies PCI_HEADER_TYPE_BRIDGE in PCIDeviceInfo,
> > > on the other hand pbc_pci_host_init() sets the register
> > > to PCI_HEADER_TYPE_NORMAL.
> > > To be honest I don't know why it does so, but that is what Blue wants.
> > 
> > BTW I think it would be prettier to have is_bridge instead of header_type
> > as a qdev property. Agree?
> 
> The spec version 3.0 defines three header types.
> 0:normal device, 1:pci-to-pci bridge, 2:card bus bridge
> So I'd like the name a bit more generic than is_bridge.
> Any suggestion?

Could we just have functions that set up header for
each type, such as
pci_init_normal_header()
pci_init_p2p_bridge_header()
pci_init_cardbus_header()

> > > So I touch only multi function bit(bit 8) and leave other bit (bit 1-7)
> > > unchanged.
> > > 
> > > If you don't like this hunk, I'll drop this hunk and leave it to Blue.
> > > What do you think?
> > 
> > Blue Swirl, could you comment on this please?
> > 
> > > static PCIDeviceInfo pbm_pci_host_info = {
> > >     .qdev.name = "pbm",
> > >     .qdev.size = sizeof(PCIDevice),
> > >     .init      = pbm_pci_host_init,
> > >     .header_type  = PCI_HEADER_TYPE_BRIDGE,	<<<<< Here
> > > };
> > > 
> > > -- 
> > > yamahata
> > 
> 
> -- 
> yamahata