The patch series passes make kernels and riscv GENERIC still boots with VirtIO-over-PCI in use, but both were done manually (unifdef apparently does not do anything to defined(A) && defined(B), only defined(A) || defined(B)) so I could easily have broken something.

\o/

This revision is now accepted and ready to land.Nov 12 2021, 5:29 PM

jrtc27 mentioned this in D32954: Retire non-NEW_PCIB code and remove config option.Jan 20 2023, 5:45 PM

love the removal of "AFU" and "gross hack"!

andrew accepted this revision.Jan 30 2023, 2:49 PM

emaste accepted this revision.Aug 3 2023, 5:52 PM

This all looks good to me... I'll see if I still have a CardBus machine around to test on... I kinda doubt it... I'll bet even if I did, it's long ago broken by something else.

sys/dev/pccbb/pccbb_pci.c
304	sc->bus can likely go away, no?
342–345	These likely can go away too.... They were part of the gross hack to allow people to set sec/sub bus numbers. This hack never worked very well...
355	This entire #if 0 likely can go... I never needed to finish stubbing it out...
sys/dev/pci/pci_pci.c
1420	Maybe this is a separate commit? IT's super-duper old hw that I doubt freebsd could still boot on... But maybe it's ok to include in this omnibus cleanup.
1435	2003/2004 era laptop... likely related to the hacks I had in cbb

Rebased. Build-tested with tinderbox -DWITHOUT_WORLDS -DMAKE_ALL_KERNELS.

This revision now requires review to proceed.Jan 26 2024, 2:01 AM

Harbormaster completed remote builds in B55630: Diff 133383.Jan 26 2024, 2:02 AM

While there is some cruft @imp noted in the ccb driver that can be purged further, that probably should be done as a separate commit.

sys/dev/pci/pci_pci.c
1420	I think in the PCI_RES_BUS case I had previously replicated this logic up in the pci(4) bus driver since it now tracks bus numbers in ivars. In particular, check `pci_reserve_secbus` in sys/dev/pci/pci.c.