On 03/08/2016 11:39 AM, Cole Robinson wrote:
Judging by how the whitelist has skewed quite far from the original
error message, I think it's better to just drop these.
If someone wants to revive this check I suggest implementing it on
a per-HV driver basis with PostParse callbacks.
Definitely the error messages for the current checks are incorrect, and
certain types that are allowed here would actually be invalid for some
hypervisors/machinetypes (the example you gave me in IRC was that vmware
doesn't have virtio-mmio, nor do most qemu machinetypes), so you are
correct that the proper place for at least some of the validation is in
a post-parse callback.
There is some value in the current validation, but it's dubious since
any failure would have led to a misleading error message. Based on that
I say ACK to this, but someone should put "improve address type
validation in postparse" on their todo list :-)