On Tue, Sep 27, 2011 at 01:39:17PM +0100, Peter Maydell wrote: > On 27 September 2011 13:32, Michael S. Tsirkin <m...@redhat.com> wrote: > > On Tue, Sep 27, 2011 at 12:50:21PM +0100, Peter Maydell wrote: > >> On 27 September 2011 12:25, Michael S. Tsirkin <m...@redhat.com> wrote: > >> > e1000 spec says CTRL.RST write should have the same effect > >> > as bus reset, except that is preserves PCI Config. > >> > Reset device registers and interrupts. > >> > > >> > Fix suggested by Andy Gospodarek <a...@greyhouse.net> > >> > >> Doesn't this have the same effect as this patch: > >> http://patchwork.ozlabs.org/patch/108673/ > > > > Right except mine clears the interrupts as well. > > I agree that's required, but why doesn't it belong in > e1000_reset() ?
I think not, because pci core resets interrupts for us. > Surely a qdev reset ought also to clear > the output irq signals... > [Disclaimer: this is fishing for somebody to explain to > me what the semantics of qdev reset actually are :-)] AFAIK what happens is that the qdev is only involved until we get to the pci root. From then on pci_bus_reset calls pci_device_reset to reset the device. > > I missed that patch - what happened to it in the end? > > I think it just got lost in the shuffle (Anthony L never > did come back and actually produce a compiler that gave > a warning on it, so I think that was just a misremembering > on his part.) > > -- PMM Yes, that just looks strange. So I intend to just stick this patch in my tree, unless someone NACKs. Then it'll get merged in due course. -- MST
