On Friday 21 January 2022 16:15:32 Martin Mareš wrote: > Hello! > > > How to handle situation when "class+subclass+prog_if" is provided and > > revision is not provided? What should libpci backends set in this case? > > Because on both Linux and Windows systems are these information provided > > separately. On Linux you can chmod 000 revision sysfs file and let class > > sysfs file still readable. Windows can probably decide itself that it > > would not report revision at all... > > Read it from the config registers in that case. On Linux "class+subclass+prog_if" can be different than what is in config registers and the purpose of this patch series with new fields is to allow user to see difference in lspci. On Windows access to real config space is not available for non-system users. > > And what to do with subsystem ids? They are not part of > > class/subclass/prog_if/revision fields and different devices have them > > stored on different locations... And for PCI-to-PCI bridges they are > > only optional and does not have to be present at all. > > I would prefer a separate PCI_FILL_xxx flag for them. So like PCI_FILL_SUBSYS flag in this patch series?
