Re: [Qemu-devel] [PATCH v6 7/7] hw/pci-bridge: format SeaBIOS-compliant OFW device node for PXB

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

On Wed, Jun 17, 2015 at 04:45:45PM +0200, Laszlo Ersek wrote:
> On 06/17/15 16:18, Kevin O'Connor wrote:
> > On Wed, Jun 17, 2015 at 03:57:36PM +0200, Michael S. Tsirkin wrote:
> >> On Wed, Jun 17, 2015 at 02:45:05PM +0200, Laszlo Ersek wrote:
> >>> SeaBIOS expects OpenFirmware device paths in the "bootorder" fw_cfg file
> >>> to follow the pattern
> >>>
> >>>   /pci-root@N/pci@i0cf8/...
> >>
> >> It's kind of crazy, isn't it?
> >> /pci@i0cf8/pci-root@N would make some sense: access rootN through cf8.
> >>
> >> But if bios needs to keep this for compatibility, maybe
> >> we have too, to. Kevin?
> > 
> > I have no issue with changing the string in SeaBIOS.  In a previous
> > email we discussed "/pci@i0cf8/pci-root@%x/" as well as
> > "/pci@i0cf8,%x/", but anything that makes sense is fine with me.
> 
> It is not fine with me.
> 
> Every time there is another idea about this format, I get to update and
> repost the OVMF series (consisting of 24 patches), which of course
> nobody on qemu-devel@ and seabios@ cares about, while it is actually the
> *only* thing that matters to me. Plus, this patch appeared in v4 and has
> been reposted without changes twice.
> 
> Honestly, the format looks outright retarded to me, but I didn't
> complain, because adopting it (and not patching SeaBIOS at all) was the
> most direct way forward. (Most direct in the sense that we're now at
> v6.) I will *not* repeat the entire discussion about the format, and I
> won't revisit that outcome. I have spent several nights and weekend days
> on implementing SeaBIOS-compatible code in qemu and OVMF, and I won't go
> back on that work.
> 
> Similarly, the patch "hw/pci-bridge: create interrupt-less, hotplug-less
> bridge for PXB" has been present in the QEMU series without functional
> changes since v2. I've been aware that it doesn't meet Michael's taste
> (that fact was documented in v2), but I'm appalled that it has taken 4
> reposts (v3 to v6) to arrive at specifics. Not only did that cause me to
> miss 4 opportunities to post an ultimately acceptable patch, it also
> wasted the reviews of Marcel and Markus, plus my work to address
> Markus's review.
> 
> I've been going out of my way to be cooperative, responsive, and just do
> whatever I've been told, minimize the impact, etc. As I said, I'm
> willing to post a v7 for the SHPC-less pci-bridge device model, but no
> more versions, and no other changes.
> 
> Thanks
> Laszlo

Sorry you feel bad.  Looks like the patches are pretty close to being
ready.

If you just address the comments about the bridge then I can merge
patches 1-5 directly.

We do need to agree about the correct paths however, this is host/guest
interface which we have to maintain forever, and it's important to get
it right. I kept hoping we can come up with something saner than
the sequence # but oh well. Do you disagree with the statement
that seabios path is currently incorrect? Kevin seems to agree.

-- 
MST