Kyösti Mälkki has posted comments on this change. ( https://review.coreboot.org/c/coreboot/+/36221 )
Change subject: WIP: Add configurable ramstage support for minimal PCI scanning ......................................................................
Patch Set 14:
Patch Set 14:
Patch Set 14:
Patch Set 14:
Patch Set 14: Code-Review+1
Are we heading towards creating all PCI APIs with and without _always like been done for read_resources()?
I believe our original goal is to block probe function for all other devices except the one marked as "mandatory" in devicetree.cb right ? @Ron
Some of us (Aaron, Nico, me) seem to be much against (that original goal or approach) of partial PCI probing/enumeration due the undesirable side-effects mentioned along the review comments here. Furthermore, some of us seem to agree the real problem is with the resource allocation and not the PCI enumeration part.
if we divide the original definition of PCI enumeration then it combination of discovering the device, marking devices numbering and allocate the resource, enabling resource. Not in static world we really don't have to do PCI device discover unless its on PCI-E root ports so resource allocation might be only significant things we have in hand right now. isn't it?
The PCI subsystem has always had separation of enumeration and resource allocation. It is only this development of "MINIMAL_PCI_SCANNING" where that was forgotten.
I think with #13 we reached some agreement that complete PCI enumeration is not harmful. I hope from now on this commit could concentrate only on addressing the somewhat urgent problem of PCI resource allocation.
As for any arguments about boot-time enhancements that could be achieved by partial PCI enumeration, let's just delay that discussion and development to another commit.