On Wed, 2018-09-12 at 13:27 +0800, Yi Min Zhao wrote:
在 2018/9/11 下午4:37, Andrea Bolognani 写道:
On Tue, 2018-09-04 at 16:39 +0800, Yi Min Zhao wrote: [...]
@@ -164,6 +164,7 @@ struct _virDomainDeviceInfo { * assignment, never saved and never reported. */ int pciConnectFlags; /* enum virDomainPCIConnectFlags */ + int pciAddressExtFlags; /* enum virDomainPCIAddressExtensionFlags */
There's a comment right above this that explains how pciConnectFlags is only used during address assignment: you should amend it to mention pciAddressExtFlags too.
As your comment on the 1st patch, if we have virPCIDeviceAddress include a extFlag, why not remove this one?
Sure, if you can get away with it that's perfect! :) However, I'm not entirely convinced you can avoid duplicating the information, because I believe there will be at least some parts of the address allocation algorithm where you'll need to access the flags but haven't figured out you're going to assign a PCI address to the device yet, which makes accessing the flags in virPCIDeviceAddress not feasible. I could very well be wrong, though: I haven't actually looked into it in detail. -- Andrea Bolognani / Red Hat / Virtualization