On Wed, Jul 20, 2022 at 02:00:16PM +0300, Roman Kagan wrote:
On Wed, Jul 20, 2022 at 11:44:26AM +0100, Daniel P. Berrangé wrote:
On Wed, Jul 20, 2022 at 01:25:55PM +0300, Roman Kagan wrote:
It's possible to create non-working configurations by attaching a device
to a derivative of PCIe slot (pcie-root-port, ioh3420, etc) and
specifying a slot number other that zero, e.g.:
-device pcie-root-port,id=s0,... \
-device virtio-blk-pci,bus=s0,addr=4,...
Make QEMU reject such configurations and only allow addr=0 on the
secondary bus of a PCIe slot.
What do you mean by 'non-working' in this case. The guest OS boots
OK, but I indeed don't see the device in the guest, but IIUC it was
said that was just because Linux doesn't scan for a non-zero slot.
Right. I don't remember if it was Linux or firmware or both but indeed
at least Linux guests don't see devices if attached to a PCIe slot at
addr != 0. (Which is kinda natural for a thing called "slot", isn't it?)
I vaguely recall there was an option to tell linux to scan all slots,
not just slot 0, not sure if that's applicable here.
That wouldn't be a broken config from QEMU's POV though, merely a
guest OS limitation ?
Strictly speaking it wouldn't, indeed. But we've had created such a
configuration (due to a bug in our management layer) and spent
non-negligible time trying to figure out why the attached device didn't
appear in the guest. So I thought it made sense to reject a
configuration which is known to confuse guests. Doesn't it?
If a configuration is a permissible per the hardware design / spec, then
QEMU should generally allow it. We don't want to constrain host side
configs based on the current limitations of guest OS whose behaviour can
change over time, or where a different guest OS may have a different POV.